A STATISTICAL ANALYSIS OF GEORGIA?S 
HMA QUALITY ASSURANCE PROCESS 
 
 
 
 
Except where reference is made to the work of others, the work described in this thesis is 
my own or was done in collaboration with my advisory committee.  This thesis does not 
include proprietary or classified information. 
 
_____________________________________ 
James Richard Willis 
 
 
 
Certificate of Approval: 
 
_______________________    __________________________ 
Rod E. Turochy     Frazier Parker, Jr. 
Assistant Professor     Professor 
Civil Engineering     Civil Engineering 
 
 
_______________________    __________________________ 
David H. Timm     Stephen L. McFarland 
Assistant Professor     Acting Dean 
Civil Engineering     Graduate School 
 
 
A STATISTICAL ANALYSIS OF GEORGIA?S 
HMA QUALITY ASSURANCE PROCESS 
 
 
James Richard Willis 
 
A Thesis 
Submitted to 
the Graduate Faculty of 
Auburn University 
in Partial Fulfillment of the 
Requirements for the 
Degree of 
Master of Science 
 
 
Auburn, Alabama 
December 16, 2005 
 
 iii
A STATISTICAL ANALYSIS OF GEORGIA?S 
HMA QUALITY ASSURANCE PROCESS 
 
 
James Richard Willis 
 
Permission is granted to Auburn University to make copies of this thesis at its discretion, 
upon request of individuals or institutions at their expense.  The author reserves all 
publication rights. 
 
 
        ________________________ 
        Signature of Author 
 
 
        ________________________ 
Date of Graduation 
 iv
VITA 
James Richard Willis, son of David Weldon and Dwina Louise (Whittle) Willis, 
was born on November 27, 1980 in San Antonio, Texas.  He graduated from Chester 
County High School as Salutatorian in 1999.  He attended Freed-Hardeman University in 
Henderson, Tennessee, for three years and then enrolled in Auburn University in August, 
2002.  In 2003, he graduated summa cum laude and with University Honors with a 
Bachelor of Science degree in Physical Science from Freed-Hardeman University.  In 
2004, he graduated summa cum laude from Auburn University with a Bachelor of Civil 
Engineering.  He immediately entered into Graduate School at Auburn University in 
August, 2004, where he worked under the supervision of the Auburn University Highway 
Research Center. 
 v
THESIS ABSTRACT 
A STATISTICAL ANALYSIS OF GEORGIA?S 
HMA QUALITY ASSURANCE PROCESS 
 
James Richard Willis 
Master of Science, December 16, 2005 
(B.C.E., Auburn University, 2004) 
(B.S., Freed-Hardeman University, 2003) 
 
117 Pages Typed 
 
Directed by Rod E. Turochy and Frazier Parker, Jr. 
 
 In the recent years, there has been a push towards allowing contractors to use their 
own test results in the quality assurance process of Departments of Transportation 
(DOTs).  While this movement saves the states money due to less testing, it has often 
been wondered if the contractor?s data compare well with GDOT data.  An analysis was 
conducted on hot mix asphalt data for the 2003 construction year in the state of Georgia 
to evaluate statistically significant differences between the contractor?s and GDOT?s data 
with the purpose of evaluating Georgia?s QA process and contractor work.  It was seen 
that while the contractor?s means compared well, significnat statistical differences were 
found with variances. 
 vi
ACKNOWLEDGEMENTS 
The author would like to thank Dr. Rod E. Turochy and Dr. Frazier Parker, Jr., for 
their guidance and support in the analysis and writing portion of this thesis.  The author 
would also like to acknowledge the National Cooperative Highway Research Program.  
This research is a small portion of the much larger Project 10-58(02). 
 vii
STYLE MANUAL AND SOFTWARE USED 
Style maunal used: MLA Handbook for Writers of Research Papers (5
th
 Edition) 
 
Computer Software Used: Microsoft Word, Microsoft Excel, Minitab 14 
 viii
TABLE OF CONTENTS 
LIST OF TABLES?????????????????????????..x 
LIST OF FIGURES?????????????????????????xiv 
CHAPTER ONE: INTRODUCTION?????????????????? 1 
1.1 Checking the Quality of Pavements?????????????????. 2 
1.2 Objectives???????????????????????????.3 
1.3 Scope????????????????????????????? 4 
CHAPTER TWO: LITERATURE REVIEW??????????????.... 6 
2.1 Terms and Definitions??????????????????????..6 
2.2 Policy?????????????????????????????10 
2.3 Sampling???????????????????????????.. 12 
2.4 Quality Assurance Analysis Methods????????????????. 13 
2.5 Studies????????????????????????????.. 15 
2.6 Summary of Findings??????????????????????... 17 
CHAPTER THREE: STATE OF THE PRACTICE????????????.. 19 
3.1 2000 University of Kentucky Survey????????????????...19 
3.2 2004 Auburn University Survey??????????????????.. 21 
3.3 GDOT Practices????????????????????????... 25 
3.4 Asphalt Content????????????????????????... 27 
3.5 Gradation???????????????????????????. 28 
 ix
3.6 Summary of Findings??????????????????????... 29 
CHAPTER FOUR: METHODOLOGY????????????????? 31 
4.1 Data Management???????????????????????? 31 
4.2 Overall Data Analysis??????????????????????. 32 
 4.2.1 The F-test???????????????????????.. 33 
 4.2.2 The Three t-tests?????????????????????34 
 4.2.3 Mean Square Deviations??????????????????36 
 4.2.4 Skewness????????????????????????37 
4.3 Individual Project Analysis????????????????????. 38 
4.4 Reduced Data Sets???????????????????????... 40 
CHAPTER FIVE: RESULTS?????????????????????41 
5.1 Overall Project Analysis?????????????????????.. 41 
5.2 Project by Project Analysis????????????????????.. 48 
 5.2.1 Statistical Analysis???????????????????? 48 
 5.2.2 Precision Tests?????????????????????.. 52 
5.3 Reduced Data Set Analysis????????????????????.. 53 
5.4 Summary of Findings??????????????????????... 54 
CHAPTER SIX: CONCLUSION???????????????????..55 
6.1 Overall Data and Reduced Data Set Analyses?????????????. 55 
6.2 Project by Project Analysis????????????????????. 57 
 x
6.4 Final Concerns and Recommendations???????????????.. 59 
REFERENCES??????????????????????????.. 61 
APPENDICES??????????????????????????... 63 
APPENDIX A: OVERALL DATA ANALYSIS SCATTERPLOTS?????... 64 
APPENDIX B: SKEWNESS ANALYSIS RESULTS???????????.. 69 
APPENDIX C: PROJECT BY PROJECT TABLES???????????? 81 
APPENDIX D: PROJECT BY PROJECT ANALYSIS SCATTERPLOTS???. 95 
 xi
LIST OF TABLES 
Table 2-1: Quality Assurance and Quality Control Properties????????.. 8 
Table 2-2: Advantages in Using CPQC?????????????????. 12 
Table 2-3: Concerns in Using CPQC??????????????????. 12 
Table 2-4: Results from Alabama Study????????????????? 16 
Table 2-5: Paired t-Test Comparisons between Kentucky Transportation Cabinet and 
Contractor Data??????????????????????????.. 17 
Table 3-1:  Materials Considered for QA/QC??????????????? 20 
 
Table 3-2:  DOT Results for Satisfaction????????????????...20 
 
Table 3-3:  Contractor Results for Satisfaction??????????????.. 21 
 
Table 3-4:  Numerical Survey Averages????????????????? 21 
 
Table 3-5: Property Responsibility??????????????????? 23 
 
Table 3-6:  Analysis Methods????????????????????? 24 
 
Table 3-7: Confidence and Satisfaction Table??????????????... 24 
 
Table 3-8: Specification Limits for Independent Samples?????????? 25 
 
Table 3-9: Allowable Percent Difference Between Department and Contractor  
 
Acceptance Tests??????????.???????????????. 27 
 
Table 5-1: F-Test Results for Contractor QCT versus GDOT QA??????... 42 
 
Table 5-2: t-Test Results for Contractor QCT versus GDOT QA??????? 42 
 
Table 5-3: F-Test Results for Contractor QCT versus GDOT Comparison???. 43 
 
 xii
Table 5-4: t-Test Results for Contractor QCT versus GDOT Comparison???. 43 
 
Table 5-5: MSD Results???????????????????????.46 
 
Table 5-6: Skewness Results?????????????????????. 47 
 
Table 5-7: Contractor QCT versus GDOT QA Variance Results by Project??? 49 
 
Table 5-8: Contractor QCT versus GDOT QA Mean Results by Project????.. 49 
 
Table 5-9: Contractor Comparison verses GDOT Comparison Variance Results by  
 
Project?????????????????????????????? 49 
 
Table 5-10: Contractor Comparison verses GDOT Comparison Mean Results 
 
by Project????????????????????????????... 49 
 
Table 5-11: Asphalt Content Precision?????????????????.. 52 
 
Table 5-12: Gradation Precision???????????????????? 53 
 
Table 5-13: Reduced Data Set Analysis for GDOT QA versus Contractor QCT?. 53 
 
Table 5-14: Reduced Data Set Analysis for GDOT Comparison versus  
 
Contractor QCT Comparison?????????????????????. 54 
 
Table 6-1:  Sample Sizes for Overall Data Results???????????? 56 
 
Table 6-2: Sample Sizes for Reduced Data Set Results??????????? 56 
 
Table 6-3: Summary of Results for Overall Data Analyses and Reduced Data Set  
 
Analyses?????????????????????????????. 56 
 
Table 6-4: QCT versus QA Results Summary??????????????... 58 
 
Table 6-5: Comparison Results Summary????????????????. 59 
 
 xiii
Table C-1: ?? Sieve Project by Project Table for QCT versus QA??????.. 82 
 
Table C-2: ?? Project by Project Table for Comparison Tests????????. 85 
 
Table C-3: #200 Sieve Project by Project Table for QCT versus QA?????... 86 
 
Table C-4: #200 Sieve Project by Project Table for Comparison Tests????? 89 
 
Table C-5: Asphalt Content Project by Project Table for QCT versus QA???... 91 
 
Table C-6: Asphalt Content Project by Project Table for Comparison Tests??? 94 
 
 
 
 
 
 
 
 
 xiv
LIST OF FIGURES 
Figure 1-1: Variance Occurences in Testing??????????????? 4 
Figure 2-1: Quality Assurance Diagram????????????????? 8 
Figure 3-1: Terminology Relationships between Georgia and TRB?????? 29 
Figure 3-2: Georgia?s Sampling Ratios?????????????????. 30 
Figure 4-1: Analysis Thought Diagram?????????????????. 36 
 
Figure 5-1: Mean Comparison from Split Sample Results for Asphalt Content?... 45 
Figure 5-2: Skewness QCT-JMF for Asphalt Content???????????.. 47 
 
Figure 5-3: Asphalt Content Variances for GDOT QA versus Contractor QCT?.. 51 
 
Figure 5-4: Asphalt Content Mean Deviations for GDOT QA versus Contractor  
 
QCT??????????????????????????????... 51 
 
Figure A-1: Mean Comparison from Split Sample Results for 1? Sieve????... 64 
Figure A-2: Mean Comparison from Split Sample Results for ?? Sieve????.. 64 
Figure A-3: Mean Comparison from Split Sample Results for ?? Sieve????.. 65 
Figure A-4: Mean Comparison from Split Sample Results for 3/8? Sieve????65 
Figure A-5: Mean Comparison from Split Sample Results for #4 Sieve????.. 66 
Figure A-6: Mean Comparison from Split Sample Results for #8 Sieve????.. 66 
Figure A-7: Mean Comparison from Split Sample Results for #50 Sieve???? 67 
Figure A-8: Mean Comparison from Split Sample Results for #200 Sieve???.. 67 
Figure A-9: Mean Comparison from Split Sample Results for Asphalt Content?.. 68 
 xv
Figure B-1: QCT-JMF Asphalt Content????????????????? 69 
 
Figure B-2: QA-JMF Asphalt Content?????????????????.. 69 
 
Figure B-3: QCT-JMF Asphalt Content (Split Sample)??????????? 69 
 
Figure B-4: DOT-JMF Asphalt Content (Split Sample)??????????? 70 
 
Figure B-5: QCT-JMF #200 Sieve??????????????????? 70 
 
Figure B-6: QA-JMF #200 Sieve???????????????????... 70 
 
Figure B-7: QCT-JMF #200 Sieve (Split Sample)????????????? 71 
 
Figure B-8: DOT-JMF #200 Sieve (Split Sample)????????????? 71 
 
Figure B-9: QCT-JMF #50 Sieve???????????????????.. 71 
 
Figure B-10: QA-JMF #50 Sieve???????????????????.. 72 
 
Figure B-11: QCT-JMF #50 (Split Sample)???????????????.. 72 
 
Figure B-12: DOT-JMF #50 Sieve (Split Sample)????????????? 72 
 
Figure B-13: QCT-JMF #8 Sieve???????????????????.. 73 
 
Figure B-14: QA-JMF #8 Sieve????????????????????. 73 
 
Figure B-15: QCT-JMF #8 Sieve (Split Sample)?????????????.. 73 
 
Figure B-16: DOT-JMF #8 Sieve (Split Sample)?????????????.. 74 
 
Figure B-17: QCT-JMF #4 Sieve???????????????????.. 74 
 
Figure B-18: QA-JMF #4 Sieve????????????????????. 74 
 
Figure B-19: QCT-JMF #4 Sieve (Split Sample)?????????????.. 75 
 
Figure B-20: DOT-JMF #4 Sieve (Split Sample)?????????????.. 75 
 
 xvi
Figure B-21: QCT-JMF 3/8? Sieve???????????????????75 
 
Figure B-22: QA-JMF 3/8? Sieve???????????????????.. 76 
 
Figure B-23: QCT-JMF 3/8? Sieve (Split Sample)????????????... 76 
 
Figure B-24: DOT-JMF 3/8? Sieve (Split Sample)????????????... 76 
 
Figure B-25: QCT-JMF ?? Sieve???????????????????.. 77 
 
Figure B-26: QA-JMF ?? Sieve???????????????????? 77 
 
Figure B-27: QCT-JMF ?? Sieve (Split Sample)?????????????. 77 
 
Figure B-28: DOT-JMF ?? Sieve (Split Sample)?????????????. 78 
 
Figure B-29: QCT-JMF ?? Sieve???????????????????. 78 
 
Figure B-30: QA-JMF ?? Sieve???????????????????? 78 
 
Figure B-31: QCT-JMF ?? Sieve (Split Sample)?????????????. 79 
 
Figure B-32: DOT-JMF ?? Sieve (Split Sample)?????????????. 79 
 
Figure B-33: QCT-JMF 1? Sieve???????????????????... 79 
 
Figure B-34: QA-JMF 1? Sieve????????????????????. 80 
 
Figure B-35: QCT-JMF 1? Sieve (Split Sample)?????????????.. 80 
 
Figure B-36: DOT-JMF 1? Sieve (Split Sample)?????????????.. 80 
 
Figure D-1: Project by Project Variances for ?? Sieve QCT versus QA????.. 95 
Figure D-2: Project by Project Average Means for ?? Sieve QCT versus QA?.....95 
Figure D-3: Project by Project Variances for ?? Sieve for Comparison Tests??. 96 
Figure D-4: Project by Project Means for ?? Sieve for Comparison Tests???.. 96 
 xvii
Figure D-5: Project by Project Variances for #200 Sieve QCT versus QA???.. 97 
Figure D-6: Project by Project Average Means for #200 Sieve QCT versus QA?. 97 
Figure D-7: Project by Project Variances for #200 Sieve for Comparison Tests?. 98 
Figure D-8: Project by Project Means for #200 Sieve for Comparison Tests??.. 98 
Figure D-9: Project by Project Variances for Asphalt Content QCT versus QA?.. 99 
Figure D-10: Project by Project Average Means for Asphalt Content QCT versus  
QA??????????????????????????????.?. 99 
Figure D-11: Project by Project Variances for Asphalt Content for Comparison  
Tests?????????????????????????????..?. 100 
Figure D-12: Project by Project Means for Asphalt Content Sieve for Comparison 
Tests????????????????????????????.??.. 100 
 
 
 
 
 
 
 
 1
CHAPTER ONE 
INTRODUCTION 
 When it comes to buying products, almost any consumer wants to get exactly 
what he or she ordered.  What would happen if a person were to visit a restaurant and 
ordered a sandwich without mayonnaise, but the bread was tarnished by only a trace of 
the unwanted condiment?  Most would not accept the sandwich and require another
one to be produced, preferably at some reduced cost.  What would a person do if they had 
ordered clothes online and upon arrival found a defect in one of the shirts?  That 
individual would most likely call the company and ask for another to be shipped to 
replace the faulty shirt.  The idea of buying quality merchandise is nothing new to the 
world of personal consumerism; however, it is also something that most Departments of 
Transportation (DOTs) also hold in high regard for their pavements.  DOTs, like 
consumers, want what they ordered, and if the contractor does not provide it within a 
specified set of limits, DOTs sometimes want new products.  Other times, the DOTs 
request reduced prices on the inferior products because the life cycle costs of the 
delivered product will be greater than that of the requested product due to more frequent 
maintenance and the possibility of a quicker replacement. 
 
 
 
 2
1.1 Checking the Quality of Pavements 
 Quality assurance is the testing process whereby a material is said to meet the 
specifications set forth for it.  It has long been the practice of DOTs to conduct specific 
tests, depending on the individual department, to determine if the material the contractor 
was using to construct the roadway in question met the specifications in the job-mix 
formula (JMF).  However, in the recent years, the DOTs have been given the option of 
relieving themselves of that burden and placing it on the shoulders of the contractors. 
 In 1995, a federal regulation entitled 23 CFR 637, Part B (Quality Assurance 
Procedures for Construction [QAPC] was enacted that allowed DOTs to begin using 
their contractor?s test results for acceptance.  In order to help ease the transition between 
testing agencies, the American Association of State Highway and Transportation 
Officials (AASHTO) published the ?Implementation Manual for Quality Assurance? and 
?Quality Assurance Guide Specification? in the year 1996.  Some states have embraced 
the idea of using contractor?s tests for material acceptance because it eases the workload 
for the DOTs.  Not only would time be saved, but also the expenses of running tests 
would be minimized if the contractors were responsible.  However, while some states 
have embraced the idea whole-heartedly, other states have been slightly more hesitant to 
turn over acceptance testing to contractors.   
 In order to investigate this concern, the National Cooperative Highway Research 
Program launched project 10-58 (02), Using Contractor-Performed Tests in Quality 
Assurance.  The objective of the project ?is to develop procedures to assist state DOTs in 
effectively using contractor-performed tests in the quality assurance process.?  To 
accomplish this goal, the results of different DOTs using contractor-performed quality 
 3
tests in the quality assurance process needed to be studied.  The methods used by each 
state were surveyed so that other suitable possible sources of data might be found to 
undergo analyzation for the stated objective, and those states that seemed to fit were 
asked to send data to be analyzed. 
 Georgia was the first state to respond to the request.  While Georgia does not use 
contractor data in its acceptance of concrete products, it does use contractor-performed 
tests on hot-mix asphalt (HMA) projects.  Both the contractor and DOTs test for the 
pavement?s asphalt content and percent passing the 1.5?, 1?, ??, ??, 3/8?, #4, #8, #50, 
and #200 sieves.  It should, however, be noted that the Georgia DOT tests for acceptance 
of mat density and smoothness. 
 
1.2 Objectives 
 The objective of this project was to numerically investigate the quality assurance 
process in the state of Georgia by the use of statistical measures and numerical 
comparisons.  Specifically, data were analyzed to determine if sets of results provided by 
the contractor differed significantly from those provided by the DOT.  Differences were 
evaluated statistically by F-tests for variance, t-tests for means, skewness, and mean 
square deviations for accuracy and precision.   
 It is not the objective of this report to prove or disprove data manipulation occurs 
on the part of either the contractor or the DOT; however, data manipulation could be a 
possible reason if contractor and DOT results are shown to have varied significantly from 
each other. 
  
 4
1.3 Scope 
 The analysis process was conducted on data collected during the year 2003 for 
Georgia?s HMA projects.  Tests were performed on both independent and split samples; 
however, while the independent samples might show differences in material uniformity, 
the split sample results analyze the differences in testing methods.  Figure 1-1 visually 
shows where variances can arise during the material testing process, and it also shows 
which variances are tested by independent and split samples. 
 
Figure 1-1. Variance Occurrences in Testing. 
 Three different data sets were analyzed: the overall data set, a reduced data set, 
and project data sets.  The overall analysis was conducted on all the data provided for the 
2003 construction season.  The nine properties previously mentioned were subjected to F 
tests, t-tests, mean square deviations, and skewness analyses to determine if the values 
the contractors were returning to GDOT were greatly different than those GDOT was 
finding from their own tests. 
 The reduced data set was compiled by combining all the data from projects that 
had at least six GDOT test results.  A large project was defined as having at least six tests 
for the purposes of this report.  F and t-tests were used on the 1/2? sieve, #200 sieve, and 
Independent 
 
Split 
 
?
2
total
 = ?
2
sampling
 + ?
2
testing
 + ?
2
material 
?
2
total 
 
?
2
testing
 (theoretically) 
 5
asphalt content data to see if many of the smaller projects might be skewing the results of 
the overall project analysis. 
 The final type of data analyzed was individual project data.  Data from all projects 
that had at least six GDOT test results were subjected to F and t-tests for the 3/8? sieve, 
#200 sieve, and asphalt content.  These tests were conducted to determine how many 
projects had significant differences in both variances and means of GDOT and contractor 
test results.  These studies strengthened observations of tendencies, such as GDOT?s 
variances being larger than the contractor?s, that might consistently occur. 
 6
CHAPTER TWO 
LITERATURE REVIEW 
 Since the policy change allowing the use of contractor data for acceptance was 
codified in the Code of Federal Regulations (CFR),  23 CFR 637, Part B (Quality 
Assurance Procedures for Construction [QAPC]), state departments of transportation 
(DOTs) have begun to use contractor data for decisions regarding payment and 
acceptance.  However, since the implementation of this policy, few states have 
investigated whether the states and contractors have adequately adapted to the change in 
responsibilities.  The slow transition from policy to practice could be due to the fact that 
each state is responsible for determining what might be appropriate for its own practice 
instead of a standard being issued by a national authority.  Other factors that might have 
influenced the lack of study could be reluctance on the part of the DOT to trust the 
contractor to adequately report the test results or the lack of understanding the terms and 
concepts involved in the process of verifying the quality of a material. 
 
2.1 Terms and Definitions  
If one looks through specification manuals or asks engineers across the country to 
define the terms related to the quality control/quality assurance process, the answers 
would be varied and inconsistent despite the efforts of organizations such as the Federal 
Highway Administration (FHWA) and the Transportation Research Board (TRB).  In 
 7
1995, when the policy change occurred in the CFR, the FHWA created a list of 
definitions of terms pertaining to the quality control/quality assurance process; however, 
to further stress the point, the TRB created the Glossary of Highway Quality Assurance 
Terms in April of 2002.  In this document, the original definitions were further refined to 
more adequately describe what the terms had come to mean.   
 Quality assurance is the first and most important term defined in either of the 
documents.  The original FHWA definition stated that QA was ?all those planned and 
systematic actions necessary to provide confidence that a product or service will satisfy 
the requirements for quality? (FHWA, 1995).  When the TRB defined this term in its 
circular, it refined the definition to the following: ?All those planned and systematic 
actions necessary to provide confidence that a product or facility will perform 
satisfactorily in service.  QA addresses the overall problem of obtaining the quality of a 
service, product, or facility in the most efficient, economical, and satisfactory manner 
possible? (TRB, 2002). 
 The term quality control (QC) is often erroneously interchanged with that of 
quality assurance.  The FHWA defined the term as ?all contractor/vendor operational 
techniques and activities performed or conducted to fulfill the contract requirements? 
(FHWA, 1995); however, QC should be viewed more in the lines of process control.  The 
TRB defined it as ?those QA actions and considerations necessary to assess and adjust 
production and construction processes so as to control the level of quality being produced 
in the end product? (TRB, 2002).   
 To further alleviate confusion between the terms, the Glossary of Highway 
Quality Assurance Terms included Figure 1 and Table 1 which have been recreated as 
 8
Figure 2-1 and Table 2-1 reproduced herein.  As can be seen, along with the many other 
differences, it should be noted that contractors are responsible for QC practices while the 
DOT is responsible for the QA process. 
Table 2-1. Quality Assurance and Quality Control Properties (TRB, 2002). 
Quality Assurance (QA) Quality Control (QC) 
? Making sure the quality of a 
product is what it should be 
? Making the quality of a product 
what it should be 
? Highway agency is responsible ? Producer/contractor is responsible 
? Includes QC ? A part of QA 
? Doing the right thing ? Doing things right 
? Motivates good QC practices ? Motivated by QA and acceptance 
procedures 
 
 
Figure 2-1. Quality Assurance Diagram (TRB, 2002). 
 As seen in Figure 2-1, there are two other legs to the QA process: acceptance and 
independent assurance.  Acceptance is simply ?the sampling and testing, or inspection, to 
determine the degree of compliance with contract requirements? (TRB, 2002).  This 
Quality Assurance in Construction 
Process Control 
(Quality Control) 
Acceptance 
Independent 
Assurance 
 9
definition is similar to the original quality control definition provided by the FHWA; 
however, the latest definitions show that acceptance relates to the contract requirements 
while QC is related to the creation of the contracted product. 
The other process, besides acceptance and QC, is the independent assurance (IA) 
testing process.  By definition, this type of testing is done by a third party, though some 
states? practices include IA tests conducted by DOT staff.  IA is defined as ?a 
management tool that requires a third party, not directly responsible for the control or 
acceptance, to provide an independent assessment of the product and/or the reliability of 
the test results obtained from the process control and acceptance testing? (TRB, 2002).  
The results of these tests are not to be used in determining if the product should be 
accepted or not. 
The FHWA simplistically defined verification as that process which was used to 
validate the quality of a product, and the newer TRB definition did not add much to the 
previous.  Verification is ?the process of determining or testing the truth or accuracy of 
test results by examining the data and/or providing objective evidence? (TRB, 2002).  
One important thing to know about verification testing is that it can be done during 
multiple phases of the quality assurance process.  Verification could be incorporated into 
the IA process, and this would be used to help verify the results of either the contractor?s 
QC tests or the agency?s acceptance tests.  The acceptance program could also be a viable 
home for a verification program ?to verify contractor testing used in the agency?s 
acceptance decision? (TRB, 2002). 
Another term pertinent to understanding the QA/QC process is pay factor or pay 
adjustment.  The TRB says that pay factors are percentages used to raise or reduce the 
 10
contractor?s payment based upon the test results that estimate the quality of the product.  
In most cases, pay factors are attributed to the product on a characteristic by 
characteristic basis.   
For example, a pavement might exceed the requirements for density inducing a 
pay factor over 100%; however, the mix might not be exceptional in its gradation causing 
the pay factor for the gradation to be below 100%.  Many agencies determine which 
characteristics to include in their payment process, and they split the percentage of the 
pay in some way across the required characteristics.  Taking the example above, the DOT 
might decide 60% of the pay should be based upon density while only 40% is based upon 
the mix?s gradation.  The two percentages are then multiplied by their respective pay 
factor, and they are then added together to produce a composite pay factor.  Another 
common approach is to take the lowest of the pay factors among all the properties and 
use that as the overall pay factor. 
 
2.2 Policy 
The FHWA created requirements for process control and acceptance programs.  
The acceptance process can include QC testing results when the following requirements 
are met (FHWA, 1995): 
1. The labs and personnel involved in the testing process are qualified. 
2. Independent samples must be used to validate the quality of the material in 
question. 
3. An independent assurance program must be in place. 
 11
The FHWA also produced a list of requirements for the acceptance programs 
including the following three items (FHWA, 1995): 
1. A frequency guide schedule 
2. Identification of sampling location 
3. Identification of attributes to be investigated. 
 
In order to accurately set up policies and specifications for using contractor-
performed quality control (CPQC) tests, a set of objectives needed to be explicitly 
outlined.  In a report submitted to the Kentucky Transportation Cabinet (KYTC), four 
distinct objectives for CPQC testing were defined (Mahboub et al., 2001). 
1. Improve the quality of the materials and processes used in the 
construction of highway projects, and reduce the life cycle costs for the 
facilities involved. 
2. Redirect the responsibility for quality control on projects to the 
contractor. 
3. Reduce the disputes between the DOT and its contractors. 
4. Enhance the construction schedule and the Department?s effort on 
quality management. 
If these objectives and policies are followed, then the use of CPQC testing should 
hold advantages for both the contractor and the DOT; however, this is not a system that 
does not raise any concerns among its users.  Tables 2-2 and 2-3 provide lists of 
advantages and concerns in the CPQC process.  These lists were compiled from surveys 
sent out in a 2001 study for the KYTC. 
 12
Table 2-2. Advantages in Using CPQC (Hancher et al., 2002). 
Agency (DOT) Contractor 
? Contractor responsible for their 
own products  
? Reduction of state personnel 
? Gaining knowledge by contractors 
? Improving dispute resolution 
? Quality improvement 
? Contractor more suitable for control 
? Improving schedule 
? Improving quality 
? Better dispute resolution 
 
Table 2-3. Concerns in Using CPQC (Hancher et al., 2002). 
Agency (DOT) Contractor 
? Validity of test data 
? Insufficient certified technician pool
? Insufficient QA 
? Lack of training 
? DOT losing expertise 
? Contractor operating at lower end of 
specification 
? Fear of losing control on project 
? Lack of understanding 
? Capability of technicians and 
facilities 
? Cost of QC 
? Lack of trust 
? Lack of training 
? Honesty of some contractors 
? Expensive independent test 
agencies 
? Different goals of contractor and 
DOTs 
 
The agency also hopes that giving this added responsibility to the contractor will increase 
the importance of quality in the contractor?s mind (Hancher et al., 2001). 
 
2.3 Sampling 
 During the sampling process, three different types of variations can be 
investigated: materials, sampling, and testing.  While the FHWA policy requires ?all 
samples used for quality control and verification sampling and testing? (FHWA, 1995) to 
be random samples, the DOT is responsible for deciding whether it shall use independent 
or split samples.  Each sample type has intrinsic characteristics that determine its 
appropriate uses. 
 13
 Independent samples are, according to the Glossary of Highway Quality 
Assurance Terms, ?taken without regard to any other sample that may also have been 
taken to represent the material in question? (TRB, 2002).  These samples are taken at 
separate times, locations, and even possibly volumes.  In turn, when independent samples 
are tested, they have the ability to provide data on the variabilities of all three parameters 
(materials, sampling, and testing) (Schmitt, 2001). 
 Split samples, unlike independent samples, come from one material source.  One 
sample is taken, and then it is broken into portions to give to the laboratories running the 
inquiries (TRB, 2002).  Due to the nature of this sampling procedure, the only variability 
should come from the testing procedures and technicians in the different laboratories.  
This is due to the samples coming from the exact same location; the materials, 
production, and sampling should all be the same (Schmitt, 2001). 
 A study conducted at the University of Wisconsin-Madison further investigated 
differences that might occur in split samples versus independent samples.  A sample of 
16 projects across six states was analyzed statistically, and the variances of the split 
samples were compared to the variances of the independent samples.  As expected, the 
results showed that the independent samples had greater variances than the split samples 
96.2% of the time (Schmitt, 2001). 
 
2.4 Quality Assurance Analysis Methods 
 Once the samples have been taken, the test results must be analyzed by using one 
of many possible methods.  While there are a variety of possibilities to choose from for 
the type of analysis to be used, two are becoming the most popular choices for state 
 14
DOTs.  The percent within limits (PWL) method is a way to compare tests results to 
specifications.  The second method, statistical analyses using F and t-tests, is a way to 
compare two test results to each other. 
 The PWL method is now highly supported by the FHWA.  In this method, the 
state DOT must decide on upper and lower limits for the chosen characteristics.  These 
limits are normally set up as two standard deviations in each direction from the mean 
using the normal distribution function.  These limits can either be chosen based upon past 
data or experience in the field.  The test results are then plotted, and the percent of test 
results within the two limits is calculated.  Payment and acceptance is based upon how 
close the test result averages are to the target and variability (Sholar et al., 2004). 
 F and t-tests are another form of statistical analysis that are commonly used to 
help analyze the consistency of contractor and DOT data.  While the theory behind these 
tests will be explained in greater detail at a later time, the purpose of each test should be 
noted.  F-tests are used to determine if the variabilities of two data sets can be shown to 
be different to a specified level of significance.  This test must be performed first to 
determine the appropriate t-test.  If the F-test shows the variabilities are the same, then a 
t-test assuming equal variances is conducted.  If the variabilities are statistically different, 
then a t-test using unequal variances is used.  T-tests are used to the test statistical 
significance of differences between the means of two data sets (Mahboub et al., 2004). 
 If the test results come from split samples, then a different type of t-test should be 
conducted.  Since the same material is being tested, the variabilities should be the same.  
The paired t-test is used on split samples because it allows a one-to-one comparison and 
is more powerfully statistically (Mahboub et al., 2004).   
 15
 While both of these statistical analyses are both commonly and easily 
incorporated into the QA programs in many states, they are not perfect, and two major 
concerns can be voiced.  The first concern is the lack of data that can be analyzed.  While 
large projects have an abundance of quality assurance testing conducted on the project, 
smaller projects might only have three or four tests.  It is difficult, if just a few tests are 
available, to accurately know whether or not the assumptions needed to conduct a proper 
t-test have been fulfilled.  The more degrees of freedom the particular test has, the more 
accurate its depiction of mean and variability similarity will be (Hancher et al., 2002). 
 The second concern arises from the first concern.  If there is a lack of data, it is 
difficult to prove normality, the test results following the normal distribution.  While 
testing and studies have shown that constructor materials test data tend to conform to a 
normal distribution, both the percent within limits method and F and t-tests rely on data 
normality (Hall et al., 2002).  If the data for some reason are not normal, then the tests 
might not accurately portray the statistical differences between the data sets (Hancher et 
al., 2002). 
 
2.5 Studies 
 A study was conducted by the Highway Research Center at Auburn University 
that compared contractor?s data to that of the Alabama DOT for a sampling of projects 
for the years 1990, 1991, and 1992.  It should be noted that this study was conducted 
before the change in the Code of Federal Regulations, but the findings are still applicable 
to this study.  When the team conducting the research specifically targeted the asphalt 
content?s difference from the job-mix formula (JMF), they found that means of contractor 
 16
and DOT measurements were different for about 1/3 of the mixes.  For the first two 
years, neither the state DOT nor the contractor?s measurements were consistently further 
from the target than the other.  However, in 1992, there were significant differences in 
means for 17 out of 48 mixes.  The state DOT?s test results consistently showed more 
deviation from the target, having the larger means for 15 of the possible 17 mixes.  
General trends observed were that the variability decreased and the accuracy increased 
with time (Parker et al, 1995). 
 After the CFR changed, another project was sponsored by ALDOT to statistically 
analyze measurements of HMA properties as the Superpave mix design system was 
implemented.  The team studied data from 1997 to 2000 for three properties: asphalt 
content, air voids, and mat density.  Using a 5% significance level on F and t-tests, 
contractor and DOT data were analyzed.  Table 2-4 provides the results which show that 
more statistical differences occurred with variances than means. 
Table 2-4. Results from Alabama Study (Parker et al., 2002).  
  Statistical Difference @ 5% 
  Variability Means 
Asphalt Content 1997 Yes Yes 
  1998 Yes No 
  1999 Yes No 
  2000 Yes No 
  Combined Yes No 
Air Voids 1997 No No 
  1998 Yes No 
  1999 Yes Yes 
  2000 Yes No 
  Combined Yes No 
Mat Density 1997 No No 
  1998 Yes Yes 
  1999 Yes Yes 
  2000 Yes Yes 
  Combined Yes Yes 
 17
 The University of Kentucky conducted a statistical analysis of the Kentucky 
Transportation Cabinet?s (KYTC) verification data versus the contractors? data.  Means 
were compared using paired t-tests for both hot-mix asphalt (HMA) and Portland cement 
concrete (PCC).   Results, as shown in Table 2-5, indicate that means of KYTC?s data set 
are comparable to the means of contractors? data.  Similar to the Alabama?s studies, the 
contractors? measurements consistently had smaller standard deviations.  This was 
explained by the possibility of using a feedback loop to correct variabilities within the 
project (Mahboub et al., 2004).  If this is truly the reasoning behind smaller contractor 
variabilities, then the contractors are doing what should be done, and the DOTs should 
not be concerned with transferring responsibilities in the QA process to them.  
Table 2-5. Paired t-Test Comparisons between Kentucky Transportation Cabinet 
and Contractor Data (Mahboub et al, 2004). 
 Significant difference at 5% 
KY testing category Mean (p-value) Standard Deviation (p-value) 
HMA-air No (0.462) Yes (<0.0001) 
HMA-asphalt content No (0.851) Yes (<0.0001) 
HMA-VMA No (0.83) No (0.854) 
PCCP-air No (0.823) Yes (0.004) 
PCCP-slump No (0.822) Yes (<0.0001) 
PCCP-strength No (0.854) Yes (0.002) 
PCC-structural-air No (0.766) No (0.219) 
PCC-structural-slump No (0.680) No (0.669) 
PCC-structural-strength No (0.480) No (0.223) 
 
2.6 Summary of Findings 
 In 1995, states were allowed to begin using contractor test results in the 
acceptance process with the implementation of 23 CFR 637 Part B.  This was done to 
help the state transportation agencies save both time and money in the testing phases of 
projects.  Since that time, studies have been conducted in Alabama and Kentucky to see 
how contractor test results compare to those of the DOT.  The 1992 Alabama study 
 18
showed that the mean asphalt contents of the contractor and DOTs were different for 
about 1/3 of the mixes tested; however, the later Alabama study and the Kentucky study 
both showed that statistically significant differences in variances were more likely to be 
found than in means.  All three studies found contractors to have smaller variances and 
mean differences from the target. 
 
 19
CHAPTER THREE 
STATE OF THE PRACTICE 
The QA process has relatively few policy requirements in terms of how it should 
be conducted.  The freedom is given to the state DOT to come up with an appropriate set 
of regulations for its contractors to follow as long as they conform to 23 CFR 637 Part B.  
The following are just a few of the questions each state must answer: 
? What properties should be considered in the QA process? 
? How many times should each property be tested for by each organization? 
? Who should conduct the tests? 
? What will the role of the contractor be? 
? What methods should be used to test each property? 
? Once the test results are in, how should it be decided if they are adequate for 
acceptance? 
With these questions and a host of others to be answered by each state, it is not 
surprising to see that possibly no two states conduct the QA process in the exact same 
way.  In fact, the diversity seen from state to state is quite staggering. 
 
3.1 2000 University of Kentucky Survey 
 In 2000, a group from the Kentucky Transportation Center at the University of 
Kentucky at Lexington sent surveys state DOTs and contractors.    The survey consisted 
 20
of questions in regards to which programs were implementing contractor data as a viable 
option for product acceptance.  Another area of concern was the satisfaction with the 
current QA practices.    Thirty state DOTs and 12 contractors responded to the survey.  
Table 3-1 is an overview of the four most common products included by DOTs in their 
2000 QA/QC specifications.  As is evident from the responses, HMA is the material for 
which contractor data is most widely used for QA purposes (Hancher et al., 2002).   
Table 3-1.  Materials Considered for QA/QC. 
Materials Percentage Using QA/QC
Grading/Earthwork 26.7% 
PCCP 50.0% 
HMA 86.7% 
Concrete Bridge Deck 50% 
 
 Tables 3-2 and 3-3 give summaries of the responses the DOTs and contractors, 
respectively.  Each DOT and contractor surveyed was given the chance to voice its 
opinion of the effects that contractor-performed acceptance testing was having in four 
categories where it was supposed to bring about positive influence: project quality, 
overall project cost, project schedule, and project disputes.  The results of this survey 
were given in number form with a 5 representing very positive and 1 representing very 
negative; however, some agencies chose not to answer, and those agencies had their 
responses marked as unidentified. 
Table 3-2.  DOT Results. 
Satisfaction 
Rating 
Project 
Quality 
Overall 
Project Cost 
Project 
Schedule 
Project 
Disputes 
Very Negative 3.4% 6.9% 0% 3.6% 
Negative 0% 17.2% 3.4% 7.1% 
No Effect 6.9% 34.5% 69.0% 21.4% 
Positive 51.7% 13.7% 3.4% 42.9% 
Very Positive 13.7% 0% 0% 0% 
Unidentified 24.1% 27.7% 24.1% 25% 
 21
 
Table 3-3.  Contractor Results. 
Satisfaction 
Rating 
Project 
Quality 
Overall 
Project Cost 
Project 
Schedule 
Project 
Disputes 
Very Negative 0% 16.7% 0% 8.3% 
Negative 8.3% 25% 0% 16.7% 
No Effect 41.7% 8.3% 33.3% 33.3% 
Positive 25% 41.7% 50% 0% 
Very Positive 25% 8.3% 16.7% 41.7% 
 
 When the averages were computed, it was found that the contractors felt more 
positive about using their test results in the categories of Overall Project Cost, Project 
Schedule, and Project Disputes.  These results are in Table 3-4.  Project Quality is the 
only area where the state DOTs felt more confident.  This seems odd that the DOT would 
feel more positive about using the contractor data than the contractors. 
Table 3-4.  Numerical Survey Averages (Hancher et al. 2002). 
Category DOT Averages Contractor Averages 
Project Quality 3.95 3.67 
Overall Project Cost 2.76 3.00 
Project Schedule 3.0 3.83 
Project Disputes 3.38 3.50 
 
3.2 2004 Auburn University Survey 
 In 2004, Auburn University?s Highway Research Center began thoroughly 
looking at the QA/QC practices of different Departments of Transportation for Project 
10-58(02) for the National Cooperative Highway Research Program.  A survey was sent 
to 25 different Departments of Transportation to investigate their quality assurance 
practices.  This survey was similar to the survey sent by the University of Kentucky, but 
questions were much more specific regarding details of the actual QA practices.  
 22
Fourteen states and Federal Highway Administration?s Western Lands Office responded 
for hot mix asphalt. 
 Table 3-5 summarizes which organization has testing responsibilities for HMA 
properties.  The number in each column represents how many agencies responded use 
what type of testing for the specified property.  This table shows that while some states 
look at similar material properties; many states choose to test more obscure properties.  
The four properties that are most consistently tested by both the contractor and the 
agency are gradation, asphalt content, mat density, and voids in mineral aggregate.  
While there may be some consistency in material properties used, the testing methods for 
a particular property may be different.   
Lot sizes vary from 500 tons maximum tonnage to one day?s work.  The 
contractors sometimes will use split samples instead of using independent samples.   
 23
Table 3-5.  Property Responsibility. 
Property Contractor Agency Both None 
Gradation 2 0 11 2 
Asphalt Content 1 1 13 0 
Voids in the Mix 0 1 7 7 
Voids in Mineral Aggregate 1 2 9 3 
Voids Filled with Asphalt 1 0 2 12 
Marshall Stability 0 0 2 13 
Flow 0 
Moisture Content 1 1 6 7 
Layer Thickness 1 4 1 9 
Mat Density 0 3 12 0 
Smoothness 3 5 6 1 
Hveem Properties 0 2 2 11 
Boil Test 0 1 0 14 
Abson Recovery 0 1 0 14 
Maximum Gravity 0 0 1 14 
Dust to Asphalt 0 0 2 13 
Retained Tensile Strength 0 0 1 14 
Fine Aggregate Angularity 0 0 1 14 
Clay Content 0 0 1 14 
Lottman 0 14 
Joint Density 0 0 2 13 
Lime Gradation 0 0 2 13 
Sand Equivalency 0 0 1 14 
Air Voids 0 0 1 14 
Mat Temperature 0 1 0 14 
 
Each agency determines the most appropriate way to decide if inconsistencies 
occur between the contractors? test results and their own.  Table 3-6 shows the variety of 
analysis methods for the four most common tests performed by both the contractor and 
the agency.  More responses would have made for a more accurate representation; 
however, the variety can still be seen from the small sampling. 
 
 24
Table 3-6.  Analysis Methods. 
Property 
Numerical 
Criteria 
F and t test t test only Other 
Gradation  
(4 responses) 
25% 50% 25% 0% 
Asphalt 
Content 
(5 responses) 
20% 60% 20% 0% 
VMA  
(3 responses) 
33.3% 33.3% 0% 33.3% 
Mat Density  
(4 responses) 
25% 25% 25% 25% 
 
 Like the survey conducted by the University of Kentucky, the Auburn survey 
asked the DOTs if they are confident the contractor?s test results provide the same control 
of quality as their test results.  This was asked for each of the properties, and then an 
overall satisfaction of the program question was posed.  Table 3-7 summarizes survey 
responses.  Only the overall and top four properties results are summarized in the table.  
The results show that, overall, the DOTs feel mostly confident using contractor test 
results in the quality assurance process. 
 
Table 3-7.  Confidence and Satisfaction Table. 
Property Confident 
Mostly 
Confident 
Neutral 
Not Totally 
Confident 
Not 
Confident 
Gradation 27.3% 36.4% 27.3% 9.1% 0% 
Asphalt 
Content 
20% 60% 10% 10% 0% 
VMA 14.3% 71.4% 14.3% 0% 0% 
Mat Density 44.4% 33.3% 22.2% 0% 0% 
Satisfaction Satisfied 
Mostly 
Satisfied 
Neutral 
Not Too 
Satisfied 
Not 
Satisfied 
Overall 33.3% 26.7% 33.3% 6.7% 0% 
 
 
 25
3.3 GDOT Practices 
The Georgia DOT accepts HMA based on four material properties: asphalt 
content, gradation, mat density, and smoothness.  However, contractors? data is used only 
for asphalt content and gradation.  Georgia DOT tests for both mat density and 
smoothness are used for acceptance. 
 The Georgia DOT?s quality assurance process terminology is different from the 
FHWA and TRB definitions provided in Chapter 2.  GDOT uses three different terms to 
designate testing methods and who did the testing.  QCT represents the regular contractor 
testing that is required, and this is related to the TRB term quality control.  Georgia 
requires one test for every 500 ton sublot, and a LOT is equal to one day?s production 
(GDOT, 2005). 
The abbreviation QA stands for the Georgia DOT?s testing that is compared to the 
QCT test results which falls under the TRB definition of acceptance.  These tests are 
conducted twice for every 5 lots or 5 days, whichever is less.  The QA tests are conducted 
on independent samples, and the results are compared to the JMF using the set of 
specification limits shown in Table 3-8.  While there are no specification limits set, the 
percents passing the 1?, 0.75?, and #50 sieves are determined (GDOT, 2005). 
Table 3-8.  Specification Limits for Independent Samples (GDOT, 2005). 
Property +/- Specification Limit
Asphalt Content 0.4% 
0.5? Sieve 6.0% 
0.375? Sieve 5.6% 
#4 Sieve 5.6% 
#8 Sieve 4/6% 
#200 Sieve 2.0% 
 
 26
 If these specification limits are met with the QA test results, then the QCT tests 
for asphalt content and gradation are permitted for use in acceptance and the calculation 
of the pay factors.  Pay factors are calculated for asphalt content, designated sieve sizes, 
and mat density.  The maximum pay factor is 1; therefore, no bonuses are given for 
exceptional work. 
 The third type of testing is DOT comparison tests.  These are on split samples 
with contractor QCT samples. The QCT and DOT Comparison tests fall under the TRB 
term independent assurance.  These split samples (Contractor QCT and DOT 
comparison) are taken once for every 10 lots, and the results are compared one to one 
with criteria in Table 3-9.  The purpose of the DOT comparison test is to verify the QCT 
results.  As with the QCT and QA samples, the percents passing the 1?, 0.75?, and #50 
sieves are determined (GDOT, 2005).   
 27
Table 3-9.  Allowable Percent Difference Between Department and Contractor 
Acceptance Tests (GDOT, 2005). 
Property Surface Mixes Subsurface Mixes 
Asphalt Content +/- 0.4% +/-0.5% 
0.5? Sieve N/A +/-4% 
0.375? Sieve +/-3.5% +/-4% 
#4 Sieve +/-3.5% +/-3.5% 
#8 Sieve +/-2.5% +/-3.0% 
#200 Sieve +/-2.0% +/-2.0% 
 
 If the DOT comparison and QCT test results compare favorably and if the DOT 
QA test results meet the specification mix requirements, QCT test results are used for pay 
factor computation.  If these conditions are not met, additional sampling and testing is 
conducted to resolve the differences.  If the differences cannot be resolved, the QCT test 
results may be replaced with the GDOT test results for pay factor computation.   
 
3.4 Asphalt Content 
 GDOT allows asphalt content to be tested by either extraction or by the ignition 
oven as specified in its specification manual.  GDT 83 is designated as the Method of 
Test for Extraction of Bitumen from Paving Mixtures using the Vacuum Extractor.  This 
test method uses a solvent and a vacuum to remove the bitumen from the HMA sample.  
The percent asphalt is calculated by then subtracting the remaining weight of the sample 
from the original weight as a percentage of the original weight (GDOT, 2005). 
 The Method of Test for Determining AC Content by Ignition Oven, GDT 125, is a 
much simpler way of determining the asphalt content of a sample.  A sample of the HMA 
is placed in an oven set to 1000
o
F.  Inside the oven is a balance that measures the weight 
of the sample.  The HMA should remain in the oven until the balance stabilizes on a 
 28
weight.  The percent asphalt is then calculated in the same way as for the extraction 
method (GDOT, 2005). 
 One reason asphalt content might have been chosen to undergo the QA process is 
its correlation to performance.  In 2004, a study conducted at North Carolina State 
University documented testing that had been conducted to characterize differences in a 
pavement?s fatigue life and initial stiffness based upon its material properties.  In testing 
asphalt content, a set of general mixes and a set of North Carolina DOT-specific mixes 
were created with asphalt content at its optimum level and at -0.5% of optimum.  On the 
North Carolina DOT mixes, the reduction in asphalt content reduced the fatigue life of 
the pavement by 18-25%; however, the general mixes showed a reduction of fatigue life 
of up to 50%.  The initial stiffness did not seem to be determined by the pavement?s 
asphalt content (Tayebali and Huang, 2004). 
 
3.5 Gradation 
 GDOT uses GDT 38, Method of Test for Mechanical Analysis of Extracted 
Aggregate, as its method for determining the gradation of the aggregate samples.  The 
sample of aggregate is sifted through a set of vertical sieves to determine the percent 
passing the specified sieves indicated earlier (GDOT, 2005). 
 While it is easy to single out a property like asphalt content and link it to 
performance, it is more difficult to single out gradation.  However, gradation is highly 
linked to the percentage of air voids in a mixture.  The finer the gradation, the lower the 
percent air voids will be.  The North Carolina State University study discussed earlier 
conducted an experiment where an SP 12.5 mm mix and an SP 19 mm mix were tested 
 29
for fatigue life.  A typical mix and a mix with a 2% increase in air voids were the two 
samples in the experiment.  A 2% increase in air voids caused the fatigue life of the SP 
12.5 mm mix to have a 40% reduction, and the increase in air voids caused the SP 19 mm 
mix to have a 60% reduction in fatigue life.  Therefore, the gradation of a mix can make a 
difference in the fatigue life of a pavement since it is tied to the air voids in the pavement 
(Tayebali and Huang, 2004). 
 
3.6 Summary of Findings 
 As shown in both the Kentucky and Auburn surveys, there are a variety of ways 
for states to organize their QA programs.  Material properties, testing methods, and 
analysis methods are just three of the differences that can be seen from state to state. 
 Terminology is another thing that varies from state to state.  While the FHWA 
and TRB set up standard definitions, GDOT?s terminology varies somewhat.  The 
relationships are show in Figure 3-1. 
 
 
Figure 3-1.  Terminology Relationships Between Georgia DOT and TRB. 
 
 Georgia specifically uses gradation and asphalt content as their properties tested 
by both the contractors and GDOT for acceptance.  Every LOT requires one QCT test by 
GDOT 
QCT 
QA 
Comparison 
TRB 
Quality Control 
 
Acceptance 
 
Independent Assurance 
 30
the contractor, and two out of every five LOTs are tested by the DOT as an independent 
sample as QA.  One out of every ten LOTs has a DOT comparison test from a split 
sample.  This is graphically shown in Figure 3-2. 
 
Figure 3.2.  Georgia DOT Sampling Ratios. 
LOTS 1           2          3          4          5          6          7          8          9         10 
QCT          1          2          3          4          5          6          7          8          9         10 
QA             1                      2                                  3                      4 
(Independent Samples) 
Comparison                                               1 
(Split Samples) 
 31
CHAPTER FOUR 
METHODOLOGY 
Two comparisons were made of contractor and GDOT test results.  The first 
consisted of comparisons of the contractor?s QCT and GDOT?s QA test results.  The 
second of comparisons of the contractor?s QCT results and GDOT?s comparison test 
results.  The properties compared were asphalt content and percent passing the 1?, 0.75?, 
0.5?, 0.375?, #4, #8, #50, and #200 sieves. 
 
4.1 Data Management 
When the Georgia DOT sent the data to Auburn University?s Highway Research 
Center to be analyzed, the numbers needed for the analyses were contained in three 
Microsoft Access databases.  Once it was understood which process each file contained, 
the organization of the column headings was not difficult to interpret. 
 The difficulties arose in retrieving correct data in the project by project analyses.  
The overall analyses were easy as the analyzer could query a specific year and material 
property using built in programs in Microsoft Access.  When the project by project 
analysis was conducted, each data row was visually inspected.  A sort might have been 
available to speed up this process; however, many of the project titles were not 
consistent.  The same project might have a ?0? (zero) instead of an ?O.?  Another project 
 32
might have been listed by the contractor as ?0100,? but GDOT listed it as ?100.?  This 
became especially difficult in trying to match projects from contractor to GDOT data. 
 This inconsistency might be due to different employees recording the data in 
different ways; however, no conclusions can be made as to why these discrepancies 
occurred, but in order to keep accurate records in the GDOT data system, identical 
project number recording is imperative.  A drop down menu providing a list of the 
projects in service at the time might be an appropriate way to help alleviate this problem. 
  
4.2 Overall Data Analysis 
 The first analysis was an overall data analysis.  All of the 2003 data were divided 
among the various properties listed above, and then these results were put through a 
series of filters to find usable data.  The first filter was one to exclude all of the data that 
did not include job mix formulas.  This was accomplished through the use of queries in 
Microsoft Access.  Only the data with job mix formulas could be used because target 
values were variable for each property; therefore, in order to combine data from job mix 
formulas, the target values had to be subtracted from the test results. 
 The second filter applied to the data was outlier removal.  Only obvious outliers, 
those impossible due to the next sieve size, could be removed from the data sets because 
there might have been inadequate mixes in the data set, and those mixes needed to be 
considered in the data analysis.  One method for finding outliers was doing a sort of the 
data in Microsoft Excel.  For example, if the results or job mix formula said 100, and the 
test results said 10 or 1 or vice versa, then the data was considered invalid, and it was 
removed.  Once the outliers were removed, actual analyses began. 
 33
4.2.1 The F-Test 
 The F-test is a test that analyzes the variances of the data sets in question.  The F-
test ?assesses mean differences by comparing the amounts of variability explained by 
different sources? (Ramsey and Shafer, 2002); therefore, the results of the test explain if 
there are differences in the variabilities of the data sets. 
 The F-tests were run for each material property in both the QCT versus QA and 
QCT Comparison versus DOT analyses using all the data passing the two removal filters 
using the data analysis programs in Microsoft Excel.  The null hypothesis was that the 
variances were not statistically different.  Two values are typically used for confidence 
levels depending on the preference of the individual performing the analyses and how 
accurate the tests need to be.  Those values are 95% and 99%.  In the case of the F tests, 
and all of the other statistical tests for this analysis, a significance level of ?=0.01 was 
used; therefore, if the p-value was less than 0.01, the variances in the two data sets were 
considered different.  This significance level was chosen to make finding differences 
between the sets more rigorous because fewer test results should be shown as different in 
a 1% significance level rather than a 5%. 
  The results of the F-tests could be important for possibly two reasons.  The 
results of the analysis would determine which type of t-test should be used for the 
remainder of the analyses for that specific material data set.  Secondly, the results could 
possibly provide insight into altering of data.  Despite the differences in the number of 
observations, the variances should still be similar; however, if one is significantly smaller 
than the other, it makes one wonder why? 
 
 34
4.2.2 The Three T-Tests 
 Different forms of t-tests provide a way to determine whether the means of two 
data sets are statistically equal.  However, in order for a t-test to be valid, at least three 
basic assumptions should be met. 
1. The distribution of the data must be normal. 
2. The two data sets must have equal standard deviations. 
3. The two data sets must be independent. 
If these requirements are not met, a variation of the basic t-test is used (Ramsey et al., 
2002). 
 Each of these assumptions is dealt with in their own distinct way.  When looking 
at normality, the Central Limit Theorem is used to justify the t-test in the case of looking 
at all of the data for the year.  The Central Limit Theorem states, ?averages based on 
large samples have approximately normal sampling distributions, regardless of the shape 
of the population distribution? (Ramsey et al. 2002).  Since the sample sizes were very 
large, it could be assumed that the distributions were normal. 
 The second assumption states that each data set should have approximately the 
same standard deviation.  This was determined using the F-test previously described.  If 
the F-test returned a p-value less than 0.01, then the hypothesis that the two variances 
were equal was void, and a t-test based on equal variances in the data sets compared was 
not applicable.  However, in this case, a t-test assuming unequal variances was used.  
There were some cases when testing the contractor QCT versus GDOT QA that this test 
was applied due to the F-test results. 
 35
 The third assumption is that the data sets are independent of each other.  While 
this was the case for the Contractor QCT versus GDOT QA test results, the QCT 
Comparison and GDOT Verification tests were conducted on split samples; therefore, 
this assumption was not met.  In this case, a paired t-test was used.  This test was a 
powerful application when it could be used, and it allowed plots to be created visually 
showing if the Contractor or GDOT had larger variances or mean discrepancies from the 
job mix formula.  Lines of absolute numerical equality were drawn on the plots to make it 
easier to tell which values were larger. 
 If all three assumptions were met, then a t-test was used for the contractor QCT 
versus GDOT QA comparison.  Figure 4-1 provides a basic decision diagram to 
determine which test was appropriate for each analysis. 
 36
 
Figure 4-1.  Analysis Thought Diagram. 
 
4.2.3 Mean Squared Deviations 
 While the F and t-tests provide individual analyses of the variances and means, 
respectively, calculating a mean squared deviation (MSD) for a data set quantifies both 
the accuracy and precision.  Three possible scenarios can be used in calculating the mean 
square deviation of a data set: larger is better, nominal is better, and smaller is better.  
Nominal is better was chosen because observations can extend from either side of the 
target.  The formula for the nominal is better scenario is expressed as Equation 1.  The 
Data 
Sets 
  Normal 
    OK? 
NO 
YES
Nonparametric 
Analysis 
Standard 
Deviation 
OK? 
Unequal 
Variances t-test 
NO
YES
Dependency 
OK? 
NO 
Paired t-test 
YES 
Basic 
t-test 
 37
nominal is better approach is the case where the results can approach from either the 
upper or the lower limit. 
22
)(
tnNIB
xxSMSD ?+=        (1) 
where: S
n
2
 = variance of data set 
 
_
x  = mean of data set 
 x
t
 = target value for data set, 0 
Smaller MSD values indicate a high level of control in the during the mix and 
construction process. 
 In this study, MSD calculations were completed for all nine material properties 
for the contractor QCT, GDOT QA, contractor comparison, and GDOT comparison data.  
The smaller values of the four show which measurements were not only more precise but 
also closer to the target job mix formula. 
 
4.2.4 Skewness 
 The final analysis conducted on the overall project data was a skewness test.  
Skewness is a way to measure the symmetry of a data set when it has been converted into 
a histogram.  If the skewness coefficient was zero, it did not necessarily mean that the 
data set was symmetrical; however, non-zero values do indicate how the data set?s 
histogram behaved.  If the value of the coefficient is negative, then the data set?s 
histogram is skewed to the left.  The converse is true for a positive skewness coefficient. 
Minitab was used to calculate skewness values.  These tests were conducted to 
analyze possible data shifts.  The Contractor?s QCT values were compared numerically to 
 38
those of GDOT?s QA and Comparison tests, but no statistical analyses, such as the 
standard error of skewness, were conducted on the actual skewness values.   
 
4.3 Project by Project Analysis 
 It was theorized that the results found for the overall data analyses might have 
been affected significantly by the large sample size; therefore, it was decided to conduct a 
further analysis for each of the large projects in Georgia?s database.  A project was 
classified as large if the smallest project data set (contractor or DOT on a particular 
project) that contained had at least six usable entries.  In the Contractor QCT versus 
GDOT QA analyses, the GDOT QA data set had to have six entries for that specific 
project, and for the QCT Comparison versus GDOT Verification analyses, six results just 
had to be recorded since split samples were used for this analysis.  The decision to define 
a large project as a project with at least six test results was arbitrarily chosen. 
 Three different properties were chosen for individual project analysis: asphalt 
content, the ?? sieve, and the #200 sieve.  These two sieves were chosen because they 
were at the opposite ends of the gradation curve.  F-tests were then run on all of the 
projects that had over six data entries to determine whether the variances of the data set 
were statistically the same at a 99% confidence level.  Once the state of variances was 
determined, the appropriate t-test was applied to the Contactor QCT versus GDOT QA 
data set.   
Paired t-tests were run on the QCT Comparison versus GDOT Verification data.  
The difference in the data sets used in the individual versus overall project analysis was 
size; therefore, the Central Limit Theorem could not be applied to use as a justification 
 39
for the normality assumption required for the appropriate use of a t-test.  However, it has 
been widely accepted that construction material property data tends to follow a normal 
distribution.  In TRB?s Synthesis of Highway Practice 232: Variability in Highway 
Pavement Construction, completed in 1996, it was assumed that construction material 
data followed a normal curve.  No other possible distribution was even considered 
(Hughes, 1996). 
Between 1996 and 1998, the Arkansas State Highway and Transportation 
Department and the University of Arkansas looked at the mix properties in asphalt 
construction in the state of Arkansas.  One analysis included normality probability plots 
to check the normality of the construction data.  The conclusion of this analysis was that 
?the data generated by the sampling and testing program executed for this project were 
found to represent a population of results that follow a normal distribution? (Hall et al., 
2002).  Since it is generally accepted that construction material properties follows the 
normal distribution, the t-test, or a variation of such a test, was a valid analysis tool for 
the data sets in question. 
When the typical test results had the job mix formula subtracted from it, the 
means and variances were calculated during the F and t-test process using Microsoft 
Excel.  Compiling these values on a project by project basis allowed for the creation of 
graphical comparisons for a specific material property for means and variances similar to 
those done in the overall project analysis with the paired t-tests.  Lines of absolute 
numerical equality were used to show if the contractor?s data or GDOT?s data were 
farther from the target in the graphs and which sets of measurements had the largest 
variability. 
 40
The project by project analysis was conducted to see if there might be just a few 
projects influencing the results of the larger databases.  However, if the majority of the 
projects showed significant differences, there might be deficiencies in the QA process in 
the state of Georgia.  The project by project analysis also allowed for analyses in dealing 
with the tendencies of contractors to either have their results higher or lower than those of 
the DOTs in both means and variances. 
 
4.4 Reduced Data Sets 
 The reduced data sets were similar in appearance to the overall project analysis; 
however, the data set consisted of a smaller sample size.  In this case, if an individual 
project was large enough to have undergone the individual analyses previously specified, 
its data were recompiled into a new database called the reduced data set.  These data 
points were analyzed by F-tests and the appropriate t-test. 
 This analysis was done to see if the larger sample size for the overall project 
analysis might have influenced the statistical tests used on the data sets.  If different 
results were found, it could possibly mean that a few smaller projects were pulling the 
overall project data away from where it should have been.  If the statistical differences 
noted for the different material properties were the same with the smaller sample size, the 
research team would feel more confident about proclaiming a statistical difference 
despite the possibility of sample size influence. 
 41
CHAPTER FIVE 
RESULTS 
As described in the methodology chapter, three different types of analyses were 
performed on the data provided by GDOT: overall project analyses, project by project 
analyses, and reduced data sets analyses.  The overall data analyses consisted of all the 
data collected during the 2003 construction season for specific material properties.  The 
project by project analyses included of all project that had at least 6 test results.  These 
projects were individually subjected to various statistical tests.  All of the data from 
individual projects that were analyzed were combined to make made up the reduced data 
set for analysis. 
 
5.1 Overall Data Analysis 
 Tables 5-1 and 5-2 provide summaries of the F and appropriate t-test results for 
comparing the Contractor QCT and GDOT QA data sets.  The tables provide numerical 
values for sample sizes, variances, means of differences from job mix formulas, and p-
values.  The tables also include if statistical differences at the 99% level were found 
between the data sets and if the property is included in pay factor calculation. 
 42
Table 5-1.  F-Test Results for Contractor QCT versus GDOT QA. 
Property n
GDOT 
S
2
GDOT, % 
n
CONT 
S
2
CONT, %
Difference p-value Pay
1? Sieve 832 1.425 4775 1.296 No 0.034 No 
?? Sieve 1637 4.167 9444 4.378 No 0.099 No 
?? Sieve 2323 6.793 13157 5.565 Yes <0.001 Yes
3/8? Sieve 2099 6.605 11587 6.044 Yes 0.004 Yes
#4 Sieve 1050 9.959 5532 7.707 Yes <0.001 Yes
#8 Sieve 2488 9.488 14051 5.534 Yes <0.001 Yes
#50 Sieve 749 4.139 4047 3.334 Yes <0.001 No 
#200 Sieve 2488 1.212 14036 0.769 Yes <0.001 No 
% Asphalt 2487 0.064 14061 0.040 Yes <0.001 Yes
 
Table 5-2.  t-Test Results for Contractor QCT versus GDOT QA. 
Property n
GDOT 
,%
_
GDOT?  
n
CONT 
,%
_
CONT?
Difference p-value Pay
1? Sieve 832 0.187 4775 0.184 No 0.941 No 
?? Sieve 1637 0.418 9444 0.535 No 0.036 No 
?? Sieve 2323 0.196 13157 0.160 No 0.530 Yes
3/8? Sieve 2099 0.246 11587 0.231 No 0.805 Yes
#4 Sieve 1050 0.320 5532 0.293 No 0.792 Yes
#8 Sieve 2488 0.253 14051 0.196 No 0.380 Yes
#50 Sieve 749 0.727 4047 0.837 No 0.170 No 
#200 Sieve 2488 0.359 14036 0.400 No 0.082 No 
% Asphalt 2487 0.004 14061 0.005 No 0.827 Yes
 
 As can be seen from Table 5-1, seven of the nine properties showed statistically 
significant differences in the variances while none of the means were shown to be 
statistically different in Table 5-2...  The statistical differences in the variances might 
stem from the sizable sample sizes used in the analyses.  The greater the sample size is, 
the more discriminating the test becomes; therefore, having sample sizes near or about 
1000 might have caused the differences in the F-tests to be so statistically profound. 
 While significant differences were shown in the statistical analyses, a simple 
visual inspection of the variances and means also was helpful in analyzing the data.  
Variances in the contractor?s data set were smaller than the variances of the GDOT data 
 43
set for eight of the nine material properties.  This might lead some to believe the 
contractors are adjusting their results, even though slightly, closer to the target value 
especially since the only property where the DOT?s variance was smaller was not used in 
pay computations.  However, while this difference shows up in the variances, only five of 
the nine contractor means are smaller for this data set, and none of the means from the 
two data sets show statistically significant differences for their means. 
 Paired t-tests were run on the data sets comprised of split samples, QCT versus 
GDOT Comparison.  Tables 5-3 and 5-4 provide the results of comparisons of variance 
and means for these data sets. 
Table 5-3.  F-Test Results for Contractor QCT versus GDOT Comparison. 
Property N
 
S
2
GDOT, % 
S
2
CONT, %
Difference p-value Pay 
1? Sieve 395 1.527 1.363 No 0.131 No 
?? Sieve 791 4.410 3.831 No 0.024 No 
?? Sieve 1067 9.343 6.576 Yes <0.001 Yes 
3/8? Sieve 953 8.479 5.545 Yes <0.001 Yes 
#4 Sieve 402 9.450 8.606 No 0.175 Yes 
#8 Sieve 1142 8.673 6.561 Yes <0.001 Yes 
#50 Sieve 282 3.971 4.004 No 0.472 No 
#200 Sieve 1141 1.137 0.791 Yes <0.001 No 
% Asphalt 1135 0.088 0.045 Yes <0.001 Yes 
 
Table 5-4.  t-Test Results for Contractor QCT versus GDOT Comparison. 
Property N
 
,%
_
GDOT?  ,%
_
CONT?
Difference p-value Pay 
1? Sieve 395 0.258 0.295 No 0.462 No 
?? Sieve 791 0.398 0.469 No 0.166 No 
?? Sieve 1067 0.314 0.118 Yes 0.002 Yes 
3/8? Sieve 953 0.516 0.329 Yes 0.005 Yes 
#4 Sieve 402 0.506 0.392 No 0.128 Yes 
#8 Sieve 1142 0.449 0.244 Yes <0.001 Yes 
#50 Sieve 282 0.897 0.763 No 0.094 No 
#200 Sieve 1141 0.334 0.447 Yes <0.001 No 
% Asphalt 1135 0.005 0.002 No 0.634 Yes 
 
 44
 One would expect fewer differences between split sample results than between 
independent sample results since the same material is being tested.  This seems to be the 
case with the variances where only five of the nine material properties show significant 
differences.  Four of these properties are used by GDOT for payment decisions.  
However, while none of the means showed statistical differences in the independent 
samples, four of the nine p-values for the split sample means fell below the required 0.01 
to be considered statistically equal. 
 Once again, the variances from the contractor?s test results were consistently 
smaller than GDOT variances.  The only property where the variance of the contractor 
was smaller was the #50 sieve, and that sieve is not included in the pay factor formula.  In 
looking at the means, the contractor?s data were closer to the target value for only five of 
the nine properties; however, those five properties are the only five properties used in the 
pay calculations. This might once again lead one to suspect a possibility of data 
manipulation on the contractor?s part. 
 To further investigate the split samples results, graphic representations were 
produced as shown in Figure 5-1 for asphalt content as an example.  All of the graphs 
produced for the overall project analysis can be found in Appendix A.  
 45
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
x=DOT Comparison-JMF (%)
Y=
Co
n
t
r
a
c
t
o
r
 Q
C
T-
JMF
 (
%
)
Linear (Lines of Absolute Numerical
Equality)
 
Figure 5-1.  Mean Comparison from Split Sample Results for Asphalt Content. 
 The graphs are divided into four quadrants by two lines of absolute numerical 
equality to show which group?s test result was the largest.  Each point in Figure 5-1 
represents the GDOT Comparison test value and the contractor?s QC test value for a split 
sample.  If the data point is contained in either of the quadrants also containing the 
horizontal axis, then the deviation from the target value for GDOT?s split sample results 
was larger than the contractor?s.  If the data point is contained in either of the quadrants 
which also contains the vertical axis, then the converse is true. 
 While many of the data points are located near the center of the graph, it can be 
seen that there are more points near the ends of the horizontal quadrants in comparison to 
the vertical quadrants.  This supports the statistical evidence that the contractor?s test 
results for asphalt content were smaller in terms of both mean and variance. 
 46
 Along with the F and t-tests, Means Squared Deviations (MSD) were calculated 
for all four data sets using the methodology described in Chapter 4.  Table 5-5 shows the 
results from the calculations.  The Contractor QCT and GDOT QA columns are the MSD 
results taken from independent samples while the Contractor QCT Comparison and 
GDOT comparison columns were calculated using the split sample test results. 
Table 5-5.  MSD Results. 
Property 
Contractor 
QCT 
GDOT 
QA 
Contractor QCT 
Comparison 
GDOT 
Comparison 
1? Sieve 1.330 1.460 1.450 1.594 
?? Sieve 4.664 4.342 4.051 4.568 
?? Sieve 5.591 6.831 6.590 9.442 
3/8? Sieve 6.097 6.666 5.653 8.745 
#4 Sieve 7.793 10.061 8.760 9.706 
#8 Sieve 5.572 9.552 6.621 8.875 
#50 Sieve 4.035 4.668 4.586 4.776 
#200 Sieve 0.929 1.341 0.991 1.249 
% Asphalt 0.040 0.064 0.045 0.088 
 
 The MSD test is used to compare both the variances and the means of the data 
sets in one term.  As can be seen, either the contractor?s comparison or QCT test results 
have the smallest MSD value for every property.  For eight of the nine properties, both of 
the contractor?s test results are smaller than either of GDOT?s values.   
One surprising trend in the data is seen when looking at the split samples MSD 
values.  For the values at the ?? sieve, 3/8? sieve, #8 sieve, and asphalt content, there 
seems to be a considerable difference in the contractor?s and DOT?s values.  However, 
these are also parameters where the variances were significantly different with a p-value 
that was less than 0.001.  This supported a hypothesis that for these data sets the variance 
would probably be the dominant parameter in the equation.  Many of the variances were 
values between 1 and 10, and all of the means were less than one.  The means were then 
 47
squared in the equation making their significance even smaller; therefore, the correlation 
to the variance data is expected. 
When using the skewness programs in Minitab, the program produces both a 
skewness coefficient and the histogram of the dataset.  Table 5-6 shows the skewness 
coefficients, and Figure 5-2 is an example of the histograms provided.  The entire set of 
histograms is in Appendix B. 
Table 5-6.  Skewness Results. 
Property Contractor QCT GDOT QA
QCT 
Comparison
GDOT Comparison
1? Sieve -0.377 -1.313 0.701 0.233 
?? Sieve 0.918 0.849 1.667 0.206 
?? Sieve 0.513 1.213 0.305 0.959 
3/8? Sieve -0.351 -0.266 0.247 0.804 
#4 Sieve 0.123 0.636 -0.204 -0.191 
#8 Sieve 0.129 0.424 0.849 0.885 
#50 Sieve 0.153 -0.480 -0.605 0.359 
#200 Sieve -0.142 1.879 -0.385 -0.156 
% Asphalt 0.105 0.969 0.444 0.593 
 
1.300.65-0.00-0.65-1.30-1.95
Skewness 0.10473
N             14061
Mean        0.00534
StDev      0.20097
 
Figure 5-2.  Skewness QCT-JMF for Asphalt Content. 
 
 When comparing the independent samples, seven of the nine skewness values 
showed the absolute value of the GDOT QA skewness coefficient to be larger than that of 
the QCT; however, for the split sample test results, only four of the nine results were 
 48
larger for the contractor test results.  It is difficult to assess the magnitude of these results 
as no standard errors of skewness were calculated.  The most significant difference was 
found in the #200 sieve between the contractor QCT and the GDOT QA.  The 
contractor?s data showed a slightly negative skew; however, GDOT?s skewness 
coefficient was highly positive.  While these two sets of data did show significance 
differences in variances with the F-test analysis, the results would not necessarily lead 
one to expect the skewness results returned. 
 
5.2 Project by Project Analysis 
5.2.1 Statistical Analysis 
 A project by project analysis was conducted for large projects on three material 
properties: ?? sieve, #200 sieve, and asphalt content.  In order for a project to be 
considered for this analysis, it had to have at least 6 records for both of the comparative 
tests.  When the data were sorted into projects, 114 projects were analyzed for the ?? 
sieves and asphalt content, and 126 projects were analyzed for the #200 sieve.  Tables 5-7 
through 5-10 provide summaries of the project-by-project results.  The column titled 
?Projects with Significantly Higher GDOT Variances? contains two percentages in it.  
The first percentage is a comparison of the projects with significantly higher GDOT 
variances with all the projects while the second number is a comparison to only the 
projects with significant differences.  The detailed summaries of the project-by-project 
analyses can be found in Appendix C. 
 49
Table 5-7.  Contractor QCT versus GDOT QA Variance Results by Project. 
Property Projects Projects with 
Larger GDOT 
Variances 
Projects with 
Significant 
Differences 
Projects with 
Significantly Higher 
GDOT Variances 
% Asphalt 114 77 (68%) 12 (10%) 10 (9%) (83%) 
?? Sieve 114 63 (55%) 13 (11%) 10 (9%) (77%) 
#200 Sieve 126 81 (64%) 17 (13%) 15 (12%) (88%) 
 
Table 5-8.  Contractor QCT versus GDOT QA Mean Results by Project. 
Property Projects Projects with 
Larger GDOT 
Means 
Projects with 
Significant 
Differences 
Projects with 
Significantly Higher 
GDOT Means 
% Asphalt 114 68 (60%) 8 (7%) 6 (5%) (75%) 
?? Sieve 114 50 (54%) 3 (3%) 3 (3%) (100%) 
#200 Sieve 126 52 (41%) 13 (10%) 6 (5%) (46%) 
 
Table 5-9.  Contractor Comparison verses GDOT Comparison Variance Results by 
Project. 
Property Projects Projects with 
Larger GDOT 
Variances 
Projects with 
Significant 
Differences 
Projects with 
Significantly Higher 
GDOT Variances 
% Asphalt 41 35 (85%) 1 (2%) 1 (2%) (100%) 
?? Sieve 34 20 (59%) 2 (6%) 2 (6%) (100%) 
#200 Sieve 45 34 (76%) 3 (7%) 3 (7%) (100%) 
 
Table 5-10.  Contractor Comparison verses GDOT Comparison Mean Results by 
Project. 
Property Projects Projects with 
Larger GDOT 
Means 
Projects with 
Significant 
Differences 
Projects with 
Significantly Higher 
GDOT Means 
% Asphalt 41 27 (66%) 1 (2%) 0 (0%) (0%) 
?? Sieve 34 15 (44%) 0 (0%) 0 (0%) (0%) 
#200 Sieve 45 21 (47%) 2 (4%) 2 (4%) (100%) 
 
 When looking at the results from the Contractor QCT and GDOT QA project by 
project analysis, it can be seen that projects where GDOT had larger means ranged from 
41% to 60%.  On the other hand, GDOT had larger variances for between 55% and 68% 
of the projects.  Twenty-four of the 354 projects had significant differences in means, and 
 50
for 15 (63%) of these, GDOT QA had the larger mean.  For variances, 42 of the 354 
projects had statistical differences, and for 35 (83%) of those, GDOT QA variances were 
the largest. 
     When looking at Tables 5-9 and 5-10, one can see that the results of the split 
sample comparisons on a project by project basis were similar to those found for the 
independent samples.  Asphalt content sieve was the only property showing a greater 
significant deviation from the target by GDOT.  GDOT values for all three properties 
were larger from 44% to 66%.  However, GDOT variances were larger between 59% and 
85% of the projects.  Only 3 of the 120 projects had significant differences between mean 
values, and two of those showed larger GDOT means.  Six of the 120 projects had 
significant differences in the variances, and all six of those projects had larger GDOT 
variances. 
 Scatterplots, as shown in Figures 5-3 and 5-4, were created to analyze the created 
data sets.  The results of the GDOT QA would be graphically represented against the 
Contractor?s QCT to show mean deviations and variances.  One of the useful results from 
a graphical approach would be the easy recognition of outliers that might appear in the 
data set where one agency?s test results were vastly different from those of the other.  All 
the created scatterplots can be found in Appendix D. 
 
 
 
 
 51
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
QA
QC
T
 
Figure 5-3.  Asphalt Content Variances for GDOT QA versus Contractor QCT. 
 
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
QA
QC
 
Figure 5-4.  Asphalt Content Mean Deviations for GDOT QA versus Contractor 
QCT. 
 52
 
5.2.2 Precision Tests 
 One of the places were variances can occur is in the test itself.  The American 
Association of State Highway and Transportation Officials (AASHTO) conducted a 
series of experiments on the testing methods themselves to determine appropriate levels 
of variation that might be explained by test itself. 
 For the measuring of asphalt content, Georgia allows its contractors to test the 
samples by either the ignition oven or the extraction method.  According to AASHTO, 
the allowable standard deviation for a multi-laboratory ignition test is 0.06%, and the 
allowable standard deviation for a multi-laboratory extraction tests is 0.29%.  In order to 
determine if the project?s test fell outside that range, a pooled standard deviation was 
calculated.  Since Georgia?s data set did not specify which test was used, the extreme 
case was used.  It was assumed that both methods were used for every project.  Table 5-
11 provides the results of the test (AASHTO, 2004).  
Table 5-11.  Asphalt Content Precision 
Test 
Method 
Independent 
Sample 
Projects 
Number of Projects 
Outside Test 
Precision 
Split 
Sample 
Projects 
Number of 
Projects Outside 
Test Precision 
Ignition 114 114 (100%) 41 41 (100%) 
Extraction 114 7 (6%) 41 10 (24%) 
 
Every project tested, whether from split or independent sample, fell outside the test 
precision for the ignition method; however, fewer fell outside the precision for the 
extraction method. 
 The precision for gradation was not dependent upon sieve size.  It was dependent 
upon the percent passing the sieve.  An average of the percent passing the sieve in 
 53
question was calculated to determine the allowable standard deviation for the test.  As 
with asphalt content, a pooled standard deviation was calculated using both the contractor 
and the GDOT data to compare to the allowable precision of the test (AASHTO, 2004).  
A summary of the results is in Table 5-12.  The results of this analysis once again show 
the test standard deviations consistently falling outside the allowable precision of the test 
for a multi-laboratory test. 
Table 5-12.  Gradation Precision 
Property Independent 
Sample 
Projects 
Number of Projects 
Outside Test 
Precision 
Split Sample 
Projects 
Number of 
Projects Outside 
Test Precision 
?? Sieve 114 101 (89%) 34 31 (91%) 
#200 
Sieve 
126 96 (76%) 45 36 (80%) 
 
5.3 Reduced Data Set Analysis 
 The third analysis was done by combining data from individual projects that were 
used for the project by project analysis into appropriate contractor or GDOT data sets.    
The data sets were compared with F and t-tests, and Tables 5-13 and 5-14 provide the 
results. 
Table 5-13.  Reduced Data Set Analysis for GDOT QA versus Contractor QCT. 
Property Projects n
G
 S
2
G
 ?
G
 n
C
 S
2
C
 ?
C
 Difference
% Asphalt 114 1410 0.058 0.011 8453 0.040 0.010 Variances 
?? Sieve 114 1385 7.701 0.146 8072 6.439 0.208 Variances 
#200 Sieve 126 1565 1.210 0.310 8908 0.741 0.367 Variances 
 
 When considering the QA versus QCT data, one can see that none of the means 
were significantly different, but all three variances were seen to be significantly different.  
These results parallel those found in the overall project analysis.  In the case of variances, 
the GDOT QA results were always larger than those for the contractor. 
 54
Table 5-14.  Reduced Data Set Analysis for GDOT Comparison versus Contractor 
QCT Comparison. 
Property Projects N S
2
GDOT
?
GDOT
S
2
CONT
?
CONT
Differences 
% 
Asphalt 
41 452 0.097 0.018 0.053 0.010 Variances 
?? Sieve 35 400 12.286 0.462 9.251 0.200 Variances 
#200 
Sieve 
45 470 7.870 0.159 7.613 0.278 Means 
 
 The results of the split sample data sets were different that those from 
independent samples.  The contractor?s variances are always smaller than those of 
GDOT, but only variances for percent asphalt and percent passing the ?? sieve were 
significantly different.  The #200 sieve had variances that were statistically similar; 
however, the means were significantly different with the contactor having the larger 
mean deviation. 
 
5.4 Summary of Findings 
 Statistically, the contractors and GDOT have similar means, but as seen in 
previous studies, the variances are where differences occur.  In all three analyses, the 
variances were more likely to be seen as statistically significant than the means were.  On 
the project level, it was seen that the majority of the projects had the contractor?s 
variances being smaller than those of GDOT.  A precision analysis was also conducted 
on project by project level to determine if the results were falling within the allowable 
standard deviations of the test.  The analysis concluded the only test having a possibility 
of fewer than 76% of the projects falling within test precision was the test to determine 
asphalt content by extraction; however, the specification test methods are not known. 
 
 55
 56
CHAPTER SIX 
CONCLUSIONS 
In doing an analysis such as the one described in this report, it is possible that 
some of the obtained results were not what was expected.  On the other hand, other 
analyses might have played directly into preconceived ideas about what the results might 
be.  The following paragraphs will provide conclusions and recommendations for data 
organization, overall and reduced project analyses, and project by project analyses. 
 
6.1 Overall Data and Reduced Data Set Analyses 
 The overall data analyses were conducted on all data that had a corresponding 
job-mix formula for the year 2003.  The reduced data set consisted of data from all 
projects that had at least six or more QA and/or Comparison tests results.  Tables 6-1 and 
6-2 provide insights into the extent of the analysis.  These data sets were compared using 
F and t-tests as described in Chapter 4.  Table 6-3 provides a brief summary of the results 
from these analyses indicating when differences in means or variances were statistically 
significant at the 99% level.  Detailed results can be found in Chapter 5. 
 57
Table 6-1.  Sample Sizes for Overall Data Results. 
 Independent Samples Split Samples
Property N
GDOT
 N
CONT
 N 
1? Sieve 832 4775 395 
?? Sieve 1637 9444 791 
?? Sieve 2323 13157 1067 
3/8? Sieve 2099 11587 953 
#4 Sieve 1050 5532 402 
#8 Sieve 2488 14051 1142 
#50 Sieve 749 4047 282 
#200 Sieve 2488 14036 1141 
% Asphalt 2487 14061 1135 
 
 
Table 6-2.  Sample Sizes for Reduced Data Set Results 
 Independent Samples Split Samples
Property N
GDOT
 N
CONT
 N 
?? Sieve 1385 8072 400 
#200 Sieve 1565 8908 470 
% Asphalt 1410 8453 452 
 
Table 6-3.  Summary of Results for Overall Data Analyses and Reduced Data Set 
Analyses. 
Overall Data Analysis Reduced Data Set Analysis Property 
QCT vs. QA QCT vs. DOT 
Comparison 
QCT vs. QA QCT vs. DOT 
Comparison 
1? Sieve No Differences No Differences   
?? Sieve No Differences No Differences   
?? Sieve Variances Variances and 
Means 
Variances Variances 
3/8? Sieve Variances Variances and 
Means 
  
#4 Sieve Variances No Differences   
#8 Sieve Variances Variances and 
Means 
  
#50 Sieve Variances No Differences   
#200 Sieve Variances Variances and 
Means 
Variances Means 
Asphalt 
Content 
Variances Variances Variances Variances 
 
 58
 As can be seen from the table, when variances were found to be statistically 
different in one analysis, the other analysis seemed to follow suit.  The elimination of 
small projects from the overall data set to the reduced data set had little impact on the 
statistical significance of differences.  The results were what were expected based upon 
the Kentucky and Alabama studies described in Chapter 2.  One theory for the majority 
of the material properties exhibiting statistical differences at the 99% confidence interval 
is the size of the samples used in the study.  The higher the number of degrees of 
freedom, the more confining the test is going to become. 
 Another theory could also possibly explain the statistical differences seen above.  
If a few of the larger projects had significant statistical differences, the results of the 
overall analysis could be swung in a favorable or unfavorable direction towards statistical 
differences. 
The split sample results show at least one statistically significant difference, 
whether for mean or variance, for 4 of the 5 pay properties.  For these five properties, 
whether a statistical difference was noted or not, the contractor?s mean value was always 
closer to the target, and its variance was always smaller than that provided by the DOT. 
 The results of the MSD and skewness analyses seemed to reiterate the findings of 
the F and t-tests.  The contractors were consistently more accurate and precise in their test 
results. 
 
6.2 Project by Project Analysis 
 In order for a project to be considered in the project analysis, it had to have at 
least six QA test results and/or Comparison tests.  These analyses were conducted on the 
 59
?? sieve, #200 sieve, and asphalt content.  The detailed results are found in Chapter 5.  
The most thought provoking data coming from these analyses were the project variance 
results.   
 While overall only a small percentage of the projects showed a statistical 
difference in the two data sets for variances, GDOT had the larger of the variances for the 
majority of the projects for all properties considered and all sample types.  One would 
think the split sample results would have been closer since they were analyzing the 
testing methods and not material properties; however, the split sample results seen in 
Table 6-5 are even more lopsided that those for the independent samples.  For the asphalt 
content projects, GDOT variances were larger for 85% of the projects. 
 Another result to consider is how many of the projects with significant differences 
had larger GDOT variances.  As said before, while there were only a few projects with 
statistical differences, over all three properties considered 83% of the projects containing 
statistical differences had larger GDOT variances for independent samples.  The split 
samples had 100% of the 6 projects with larger GDOT variances if a statistical difference 
was noted.  Tables 6-4 and 6-5 summarize the individual project analyses. 
Table 6-4.  QCT versus QA Results Summary. 
Property Projects Projects 
with 
Larger 
GDOT 
Variances 
Projects 
with 
Significant 
Variance 
Differences 
Projects 
with 
Significantly 
Higher 
GDOT 
Variances 
Projects 
with 
Larger 
GDOT 
Means 
Projects 
with 
Significant 
Mean 
Differences 
Projects 
with 
Significantly 
Higher 
GDOT 
Means 
?? Sieve 114 63 (58%) 13 (11%) 10 (9%) 50 
(54%) 
3 (3%) 3 (3%) 
#200 
Sieve 
126 81 (64%) 17 (13%) 15 (12%) 52 
(41%) 
13 (10%) 6 (5%) 
% 
Asphalt 
114 77 (68%) 12(10%) 10 (9%) 68 
(60%) 
8 (7%) 6 (5%) 
 
 60
Table 6-5.  Comparison Results Summary. 
Property Projects Projects 
with 
Larger 
GDOT 
Variances 
Projects 
with 
Significant 
Variance 
Differences 
Projects 
with 
Significantly 
Higher 
GDOT 
Variances 
Projects 
with 
Larger 
GDOT 
Means 
Projects 
with 
Significant 
Mean 
Differences 
Projects 
with 
Significantly 
Higher 
GDOT 
Means 
?? Sieve 34 20 (59%) 2 (6%) 2 (6%) 15 
(44%) 
0 (0%) 0 (0%) 
#200 
Sieve 
45 34 (76%) 3 (7%) 3 (7%) 21 
(47%) 
2 (4%) 2 (4%) 
% 
Asphalt 
41 35 (85%) 1 (2%) 1 (2%) 27 
(66%) 
1 (2%) 0 (0%) 
 
 One would think these percentages for both the independent samples and the split 
samples would be closer to 50%, but for some reason, the contractor?s numbers seem to 
be consistently closer to the value specified in the JMF.  This might be due to improved 
testing ability with frequent testing, or it might be a symptom of a larger problem. 
 
6.3 Final Concerns and Recommendations 
? A data management system should be set up to control the proper recording of 
project numbers.  This could be accomplished by either having one person input 
all of the data for one project or by having a menu to select projects from. 
? The properties selected by the state of Georgia to incorporate into their QA plan 
could possibly be amended.  At the present time, contractor testing is only done 
on mix properties while GDOT tests for mat density and smoothness.  Many 
states incorporate a contractor-performed mat density test which might be useful 
in making sure the product is laid and compacted properly.  Another concern 
might arise in the number of sieves incorporated into the pay factor.  If one of the 
upper sieves is collects too high or too low a percentage of the aggregates, then 
 61
the corresponding sieves below will also be off in a compounding nature.  One 
mistake in an upper sieve might prove harmful to the actual pay of the project 
while not being very detrimental to the performance of the mix.  Georgia has four 
consecutive sieves that are used for pay factor.  This probably leads to a high 
correlation between the sieves.  Spreading out the sieves used in the pay factor 
computation would decrease the correlation and might lead to a better 
representation of the mix. 
? During the precision analysis, it was seen that the standard deviations for the 
projects were consistently falling outside the allowable standard deviation based 
upon precision tests conducted on the testing method.  The extraction method was 
the only test that did not have 75% or more of its tests falling outside the 
allowable standard deviation.  More tests failed to fall within precision limits than 
contained statistically significant differences.  
? The biggest concern with the Georgia DOT data at this time is the variance results 
reported.  At the present time, Georgia accepts data based upon a percentage 
value, but no statistical analysis is completed.  Some states conduct F and t-tests 
in order to determine if the material is acceptable or not.  It might be worthwhile 
for Georgia to use these statistical tests to keep a closer eye on its quality 
control/quality assurance process.  If not these statistical analyses, then some 
other means of monitoring more than just the means should be employed. 
 62
REFERENCES 
AASHTO.  Standard Specifications for Transportation Materials and Methods of 
Sampling and Testing.  AASHTO: Washington, D.C., 2004. 
FHWA. 23 CFR Part 637.  Washington D.C., 1995. 
GDOT.  GSP 21.  <http://tomcat2.dot.state.ga.us/thesource/pdf/gsp/gsp21.html>.  
October 21, 2005. 
Hall, Kevin D. and Stacy G. Williams.  ?Establishing Variability for Hot-Mix Asphalt 
Construction in Arkansas.?  Journal of the Transportation Research Board 1813: 
2002. 
Hancher, Donn E., Yuhong Wang, and Kamyar C. Mahboub.  Contractor Performed 
Quality Control on KyTC Projects.  Kentucky Transportation Center: Lexington, 
Kentucky: 2002. 
Mahboub, Kamyar C., Donn E. Hancher, and Yuhong Wang.  ?Contractor-Performed 
Quality Control: Is the Fox Guarding the Henhouse??  Journal of Professional 
Issues in Engineering Education and Practice Volume 130:4.  ASCE: October 
2004. 
Mahboub, Kamyar C. and Donn E. Hancher.  Development of Concrete QC/QA 
Specifications for Highway Construction in Kentucky.  Kentucky Transportation 
Center: Lexington, Kentucky, 2001. 
 63
Parker, Frazier, Jr. and Md. Shabbir Hossain.  ?Hot-Mix Asphalt Mix Properties 
Measured for Construction Quality Control and Assurance.?  Journal of the 
Transportation Research Board 1469: 1995.
Parker, Frazier, Jr. and M. Shabbir Hossain.  ?Statistics for Superpave Hot-Mix Asphalt 
Construction Quality Control/Quality Assurance Data.?  Journal of the 
Transportation Research Board 1813: 2002. 
Ramsey, Fred L. and Daniel W. Schafer.  The Statistical Sleuth.  Pacific Grove, 
California: 2002.
Schmitt, Robert L., Awad S. Hanna, Jeffrey S. Russel, and Erik V. Nordheim.  ?Analysis 
of Bias in HMA Field Split-Sample Testing.?  Journal of the Association of 
Asphalt Paving Technologists. Volume 70: 2001. 
Sholar, Gregory A., Gale C. Page, James A. Musselman, Patrick B. Upshaw, and Howard 
L. Moseley.  ?Development of the Florida Department of Transportation?s Percent 
Within Limits Hot-Mix Asphalt Specification.?  TRB 84
th
 Annual Meeting 
Compendium of Papers CD-ROM: 2005. 
Tayekabli, Akhtarhusein and Yuanxiong Huang.  Material Characterization and 
Performance Properties of Superpave Mixtures.  FHWA: Washington, D.C., 
2004. 
Transportation Research Board.  Glossary of Highway Quality Assurance Terms. 
Washington D.C., 2002. 
 64
APPENDICES 
 
 65
APPENDIX A 
OVERALL DATA ANALYSIS SCATTERPLOTS 
-8
-6
-4
-2
0
2
4
6
8
-8 -6 -4 -2 0 2 4 6 8
x = DOT-JMF (% Passing)
y = Q
C
T
-
JM
F (
%
 Pas
s
i
n
g)
Linear (Line of
Absolute Numerical
Equality)
 
Figure A-1.  Mean Comparison from Split Sample Results for 1? Sieve. 
-20
-15
-10
-5
0
5
10
15
20
-20-15-10-5 5 101520
x = DOT-JMF (% Passing)
y
 =
 Q
C
T-
J
M
F
 
(
%
 
P
a
ssi
n
g
)
Linear (Line of
Absolute Numerical
Equality)
 
Figure A-2.  Mean Comparison from Split Sample Results for 3/4? Sieve. 
 66
-25
-20
-15
-10
-5
0
5
10
15
20
25
-25 -20 -15 -10 -5 0 5 10 15 20 25
x = DOT-JMF (% Passing)
y =
 
Q
C
T
-
J
M
F (%
 Pas
s
i
ng
)
Linear (Line of Absolute
Numerical Equality)
 
Figure A-3.  Mean Comparison from Split Sample Results for 1/2? Sieve. 
 
-15
-10
-5
0
5
10
15
-15 -10 -5 0 5 10 15
x = DOT-JMF (% Passing)
y = 
QC
T
-
JM
F (% Pas
s
i
ng
)
Linear (Line of Absolute Numerical
Equality)
 
Figure A-4.  Mean Comparison from Split Sample Results for 3/8? Sieve. 
 67
 
Figure A-5.  Mean Comparison from Split Sample Results for #4 Sieve. 
-25
-20
-15
-10
-5
0
5
10
15
20
25
-25 -15 -5 5 15 25
x = DOT-JMF (% Passing)
y
 = Q
C
T
-
J
M
F (
%
 Pa
s
s
i
n
g
)
Linear (Line of Absolute
Numerical Equality)
 
Figure A-6.  Mean Comparison from Split Sample Results for #8 Sieve. 
-15
-10
-5
0
5
10
15
-15 -10 -5 0 5 10 15
x = DOT-JMF (% Passing)
y =
 QCT
-
JM
F
 (
%
 
Pas
s
i
ng)
Linear (Line of Absolute
Numerical Equality)
 68
-10
-8
-6
-4
-2
0
2
4
6
8
10
-10 -5 0 5 10
x = DOT-JMF (% Passing)
y
 = Q
C
T
-
J
M
F (
%
 Pa
s
s
i
n
g
)
Linear (Line of
Absolute Numerical
Equality)
 
Figure A-7.  Mean Comparison from Split Sample Results for #50 Sieve. 
-6
-4
-2
0
2
4
6
-6 -4 -2 0 2 4 6
x = DOT-JMF (% Passing)
y = 
QC
T
-
JM
F (%
 Pa
s
s
i
ng)
Linear (Line of Absolute
Numerical Equality)
 
Figure A-8.  Mean Comparison from Split Sample Results for #200 Sieve. 
 69
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
x = DOT-JMF (Aspalt Content)
y
 = QC
T
-
J
M
F (As
pha
l
t
 C
o
nte
n
t)
Linear (Line of
Absolute Numerical
Equailty)
 
Figure A-9.  Mean Comparison from Split Sample Results for Asphalt Content. 
 70
APPENDIX B 
SKEWNESS ANALYSIS RESULTS 
1.300.65-0.00-0.65-1.30-1.95
Skewness 0.10473
N             14061
Mean        0.00534
StDev      0.20097
 
Figure B-1.  QCT-JMF Asphalt Content 
1.300.65-0.00-0.65-1.30-1.95
Skewness 0.9689
N             2487
Mean       0.00416
StDev      0.25333
 
Figure B-2.  QA-JMF Asphalt Content 
1.81.20.60.0-0.6-1.2-1.8
Skewness 0.44370
N             1135
Mean        0.00211
StDev      0.21171
 
Figure B-3.  QCT-JMF Asphalt Content (Split Sample) 
 71
1.81.20.60.0-0.6-1.2-1.8
Skewness 0.59261
N             1135
Mean       0.00547
StDev      0.29696
 
Figure B-4.  DOT-JMF Asphalt Content (Split Sample) 
6.44.83.21.60.0-1.6-3.2-4.8
Skewness -0.142382
N              14306
Mean         0.39994
StDev        0.87716
 
Figure B-5.  QCT-JMF #200 Sieve 
6.44.83.21.60.0-1.6-3.2-4.8
Skewness 1.8794
N              2488
Mean        0.3594
StDev      1.1011
 
Figure B-6.  QA-JMF #200 Sieve 
 72
4.53.01.50.0-1.5-3.0-4.5-6.0
Skewness -0.38469
N             1141
Mean       0.44722
StDev     0.88951
 
Figure B-7.  QCT-JMF #200 Sieve (Split Sample) 
4.53.01.50.0-1.5-3.0-4.5-6.0
Skewness -0.15567
N             1141
Mean       0.33364
StDev      1.06621
 
Figure B-8.  DOT-JMF #200 Sieve (Split Sample) 
10.57.03.50.0-3.5-7.0-10.5-14.0
Skewness 0.152838
N             4047
Mean       0.8367
StDev     1.8258
 
Figure B-9.  QCT-JMF #50 Sieve 
 73
10.57.03.50.0-3.5-7.0-10.5-14.0
Skewness -0.48027
N              749
Mean        0.7274
StDev       2.0343
 
Figure B-10.  QA-JMF #50 Sieve 
9630-3-6-9
Skewness -0.60465
N               282
Mean         0.7628
StDev       2.0010
 
Figure B-11.  QCT-JMF #50 Sieve (Split Samples) 
9630-3-6-9
Skewness 0.358621
N              282
Mean        0.89681
StDev       1.99264
 
Figure B-12.  DOT-JMF #50 Sieve (Split Samples) 
 74
24181260-6-12
Skewness 0.12858
N              14051
Mean        0.1958
StDev       2.3525
 
Figure B-13.  QCT-JMF #8 Sieve 
24181260-6-12
Skewness 0.4244
N              2488
Mean        0.2528
StDev      3.0802
 
Figure B-14.  QA-JMF #8 Sieve 
20151050-5-10-15
Skewness 0.84939
N             1142
Mean        0.2439
StDev      2.5614
 
Figure B-15.  QCT-JMF #8 Sieve (Split Samples) 
 75
20151050-5-10-15
Skewness 0.88454
N              1142
Mean       0.4494
StDev       2.9451
 
Figure B-16.  DOT-JMF #8 Sieve (Split Samples) 
22.016.511.05.50.0-5.5-11.0
Skewness 0.12328
N              5532
Mean       0.2930
StDev      2.7762
 
Figure B-17.  QCT-JMF #4 Sieve 
22.016.511.05.50.0-5.5-11.0
Skewness 0.63557
N             1050
Mean        0.3205
StDev      3.1558
 
Figure B-18.  QA-JMF #4 Sieve 
 76
1284-0-4-8-12
Skewness -0.20406
N             402
Mean        0.3820
StDev       2.9335
 
Figure B-19.  QCT-JMF #4 Sieve (Split Samples) 
1284-0-4-8-12
Skewness -0.19096
N              402
Mean        0.5065
StDev       3.0740
 
Figure B-20.  DOT-JMF #4 Sieve (Split Samples) 
15105-0-5-10-15
Skewness -0.35097
N              11587
Mean        0.2312
StDev       2.4585
 
Figure B-21.  QCT-JMF 0.375? Sieve 
 77
15105-0-5-10-15
Skewness -0.26617
N               2099
Mean         0.2461
StDev       2.5700
 
Figure B-22.  QA-JMF 0.375? Sieve 
20151050-5-10
Skewness 0.24666
N              953
Mean        0.3285
StDev      2.3548
 
Figure B-23.  QCT-JMF 0.375? Sieve (Split Samples) 
20151050-5-10
Skewness 0.80359
N              953
Mean        0.5162
StDev       2.9118
 
Figure B-24.  DOT-JMF 0.375? Sieve (Split Samples) 
 78
28211470-7-14
Skewness 0.51317
N             13157
Mean        0.1596
StDeva     2.3591
 
Figure B-25.  QCT-JMF 0.5? Sieve 
28211470-7-14
Skewness 1.2129
N              2323
Mean        0.1960
StDeva    2.6063
 
Figure B-26.  QA-JMF 0.5? Sieve 
181260-6-12-18
Skewness 0.30511
N             1067
Mean       0.1176
StDeva    2.5644
 
Figure B-27.  QCT-JMF 0.5? Sieve (Split Samples) 
 79
181260-6-12-18
Skewness 0.95880
N              1067
Mean        0.3137
StDev       3.0566
 
Figure B-28.  DOT-JMF 0.5? Sieve (Split Samples) 
10.57.03.5-0.0-3.5-7.0-10.5-14.0
Skewness 0.91821
N              9444
Mean        0.5348
StDev      2.0923
 
Figure B-29.  QCT-JMF 0.75? Sieve 
10.57.03.5-0.0-3.5-7.0-10.5-14.0
Skewness 0.84905
N              1637
Mean        0.4176
StDev      2.0413
 
Figure B-30.  QA-JMF 0.75? Sieve 
 80
12840-4-8
Skewness 1.66709
N              791
Mean        0.4688
StDev       1.9574
 
Figure B-31.  QCT-JMF 0.75? Sieve (Split Samples) 
12840-4-8
Skewness 0.2056
N              791
Mean       0.3979
StDev      2.1000
 
Figure B-32.  DOT-JMF 0.75? Sieve (Split Samples) 
6420-2-4-6-8
Skewness -0.37664
N               4775
Mean        0.18421
StDev       1.13825
 
Figure B-33.  QCT-JMF 1? Sieve 
 81
6420-2-4-6-8
Skewness -1.3132
N               832
Mean         0.1874
StDev       1.1937
 
Figure B-34.  QA-JMF 1? Sieve  
420-2-4-6
Skewness 0.70112
N             395
Mean       0.29468
StDev      1.16758
 
Figure B-35.  QCT-JMF 1? Sieve (Split Samples) 
420-2-4-6
Skewness 0.23251
N             395
Mean       0.25848
StDev     1.23553
  
Figure B-36.  DOT-JMF 1? Sieve (Split Samples) 
 82
APPENDIX C 
PROJECT BY PROJECT TABLES 
 83
Table C-1.  ?? Sieve Project by Project Table for QCT versus QA. 
Project # Nqc Nqa Vqc Vqa Mqc Mqa Differences
BRN-42-2(40)01 13 6 1.7856 1.2457 -1.1308 -1.1167 None 
CM-186-1(28)01 38 7 11.1322 13.2024 0.6421 0.9286 None
CSSTP-M002-
00(444)01 27 6 0.2972 1.9417 2.5519 2.4167 Variances 
CSSTP-M002-
00(453)01 54 7 0.0007 0.0000 -0.0037 0.0000 None 
EDS-27(136)01 138 35 4.8818 28.2736 0.3449 1.6229 Variances 
EDS-27(147)01 90 14 2.3432 1.5782 -1.1922 -1.0857 None 
EDS-84(16)01 127 16 7.3232 5.3740 0.6583 0.5750 None
EDS-84(17)01 64 11 6.4686 9.9626 0.6000 0.2636 None 
EDS-545(8)01 71 9 7.6223 4.2553 -0.4099 0.0556 None
EDS-555(6)01 73 15 7.8346 12.1884 -0.0603 1.0533 None 
EDS-565(7)-01 99 7 7.8370 13.9381 0.4657 0.4143 None
EDS-565(15)-01 98 15 5.7659 8.2995 0.8276 0.7333 None 
FLF-540(8)01 61 8 5.8679 4.1250 1.0672 1.2750 None
FLF-540(25)01 95 12 3.6104 3.6924 1.4168 0.9167 None 
GIP-341(33)01 107 13 9.2501 9.6309 0.8533 0.6385 None
GIP-341(34)01 146 25 14.9424 27.4157 0.3671 -0.7640 None 
GIP-341(38)01 203 21 14.0092 10.8481 0.9163 1.1714 None
HPP-3717-00(300)01 28 16 0.0714 0.1873 -0.3786 -0.3937 None 
HPP-STP-178-1(33)01 23 7 2.1225 0.9657 0.4391 0.3714 None 
IM-00MS(268)01 41 14 4.1352 8.1385 -0.2317 -0.4000 None
IM-285-1(349)01 26 7 13.5448 113.5148 -0.7346 1.3857 Variances 
IM-0000-00(471)01 54 15 10.2602 6.1898 1.0593 1.5467 None 
IM-0000-00(470)01 87 11 10.2481 13.3836 1.4839 1.3182 None 
IM-00MS(20)01 56 17 5.4579 8.8272 1.3750 1.4706 None
IM-MS(325)01 31 7 5.9792 2.8800 -1.5484 -2.1000 None 
IM-NH-85-2(148)01 175 20 10.5233 11.2510 1.1863 1.0950 None 
LAR00-S005-
00(593)C1 13 8 0.4990 0.4536 0.2308 0.3250 None
LAR00-S005-
00(649)C1 13 6 0.0000 0.0067 0.0000 -0.0333 Variances 
LAR31-87-1(067)01 23 7 0.0157 0.0057 -0.0261 -0.0286 None 
LAR32-2-3(067)01 17 7 0.3368 0.3262 -0.3059 -0.3571 None
LAR32-2-5(63)C1 12 6 0.0000 0.0150 0.0000 -0.0500 Variances 
LAR32-24-1(057)C1 42 8 0.7603 0.4570 -0.3214 0.0625 None 
LAR32-30-2(89)01 34 9 29.7617 28.5753 0.9206 -2.5556 None
LAR32-250-1(057)01 37 8 0.6825 0.5998 -0.5973 -0.2625 None 
LAR32-888-1(089)01 35 11 3.7732 7.0876 0.4971 1.7182 None 
LAU32-8532-69(121)01 28 10 1.0177 0.0573 -0.6929 -0.2200 Variances 
NH-8013(8)01 63 20 8.3353 4.9237 1.7175 0.4450 None 
NH-11-1(44)01 32 11 11.3519 2.8787 1.7188 -0.8545 None
NH-012-1(83)01 60 10 11.4486 17.9107 0.6417 0.1800 None 
NH-017-2(53)01 36 12 10.8882 16.4717 -0.3583 -0.2417 None
NH-20-2(183)01 193 24 7.0404 8.6573 0.5041 2.3083 Mean 
 84
NH-026-2(81)01 29 9 7.7946 8.5050 0.6034 -0.2333 None 
NH-043-1(49)01 14 6 9.5191 4.6387 -0.5286 -0.6667 None
NH-56-1(62)01 43 16 15.4562 17.5320 1.6372 -0.6563 None 
NH-75(157)01 234 30 10.0084 7.7081 0.3910 -0.2767 None
NH-75-1(203)01 180 17 9.4872 17.3468 1.1406 1.1059 None 
NH-171-1(4)01 100 8 4.4169 2.1943 2.1360 1.3000 None
NH-IM-75-1(158) 174 24 9.0923 8.5748 -0.9506 -0.9208 None 
NH-IM-95-1(119)01 90 13 9.5126 23.0640 -0.2078 -1.7308 Variances 
NH-IM-95-1(125)01 141 29 4.7353 5.3664 -0.6099 -0.5000 None 
NH-IM-95-1(155)01 71 11 15.2213 9.6702 0.7423 3.5273 None 
NHS-0000-00(768)01 61 8 7.8002 11.5813 2.0426 3.4125 None 
NHS-M002-00(398)01 487 53 5.3242 5.2975 0.0031 -0.0434 None 
NHS-M001-00(530)01 192 9 13.0622 11.9128 1.5516 1.6556 None 
NHS-M001-00(531)01 59 12 1.8749 3.7355 -0.8305 -0.5500 None 
NHS-M001-00(591)01 50 13 3.5058 3.4500 -1.5560 -0.2000 None 
NHS-M002-00(276)01 23 8 2.0715 3.9200 1.2652 0.9000 None 
NHS-M002-00(292)01 247 10 5.9524 9.4250 -0.0142 -0.0180 None 
PR000-S005-
00(810)C1 18 7 0.3657 0.0129 0.1278 -0.0429 Variances 
PR-147-2(055)01 8 7 0.0000 0.0000 0.0000 0.0000 None 
PR131-1(63)C1 8 6 0.0000 0.0000 0.0000 0.0000 None
PR-1963-2(139)01 28 6 7.1455 12.4910 1.7536 2.0500 None 
PR-8530-83(39)C1 22 10 1.1089 5.3044 0.9682 0.6000 Variances 
PR-8531-21(77)C1 51 11 12.8023 14.1076 0.9078 0.6818 None 
PRLOP8530-32(15)C1 31 6 5.4603 3.7147 0.0032 -0.4333 None 
SAMA-60SP(4)01 30 6 2.8886 5.6800 0.2800 1.2000 None
SAMA-3(249)01 38 9 6.5743 3.4344 -0.1974 2.5222 Mean 
SAMA-3(294)01 123 25 1.1791 1.0867 -0.7772 -0.8400 None 
SAMA-3(298)01 198 31 2.3558 1.3203 0.3535 0.4677 None
SAMA-39(44)01 55 14 4.4110 0.9455 -0.4709 -0.3643 Variances 
SAMA-53(133)01 34 6 0.6020 0.6787 0.3529 0.3333 None 
SAMA-74(56)01 25 9 1.2208 0.8700 1.5920 0.8333 None
SAMA-74(58)01 121 23 1.2860 0.8421 -1.2281 -1.2130 None 
SAMA-155(52)01 18 6 0.7662 2.3680 -1.1167 -2.8000 Mean 
SAMA-206-C0(3)01 45 7 0.6860 1.0795 0.1244 -0.0429 None 
SAMA-234-(17)01 51 8 0.6792 0.9484 2.0608 1.8625 None
SAMA-520(58)01 84 18 0.4159 0.5603 0.3798 0.4444 None 
SAMA-520(59)01 144 20 0.7411 0.7887 -0.3514 -0.3850 None
SAM-M002-00(405)01 40 6 0.4058 0.8750 0.7325 0.7500 None 
SAM-M002-00(417)01 46 7 0.6215 0.5690 0.5065 0.1286 None 
SAM-M002-00(422)01 64 14 1.8148 1.1352 1.4016 1.8857 None 
SAM-M002-00(399)01 23 7 2.6650 1.5057 0.4957 0.3714 None 
STP-001-5(60)01 37 9 0.6994 0.7361 0.5000 0.5111 None
STP-9-2(78)01 18 8 0.5776 2.5255 0.4667 0.8375 Variances 
STP-9-2(90)01 21 8 3.7786 2.7927 -0.0190 0.2125 None 
STP-34-1(22)01 89 14 2.5506 2.8623 -1.8416 -0.7071 None
STP-36-1(13)01 49 15 10.8128 9.9617 0.5367 -0.1200 None 
STP-042-2(49)01 82 8 11.3187 21.1070 -0.0171 -2.4875 None
 85
STP-141-1(7)01 36 21 7.9551 10.5381 0.6972 -0.0286 None 
STP-149-1(29)01 62 8 6.6796 9.2314 0.2065 0.9000 None
STP-803(4)01 72 18 7.2362 5.5335 -0.2694 -0.6056 None 
STP-2434(2)01 14 6 0.7495 3.9710 0.6786 1.0500 Variances 
STP-M00-00(703)01 34 7 0.4411 0.2733 1.8794 1.7000 None 
STP-M001-00(273)01 82 19 1.4091 1.2050 -0.8329 -0.2053 None 
STP-M001-00(330)01 88 14 1.9975 0.8569 -1.2477 -0.9000 None 
STP-M001-00(490)01 25 6 2.4800 3.0297 0.7800 0.5167 None 
STP-M001-00(492)01 50 11 0.5925 0.6945 -0.8340 -0.8364 None 
STP-M001-00(493)01 97 10 3.0188 2.4827 0.4784 -0.0600 None 
STP-M001-00(773)01 76 8 2.1427 3.5621 -0.0408 0.3750 None 
STP-M001-00(872)01 77 6 1.7208 0.4867 -0.3909 0.2333 None 
STP-M002-00(093)01 45 7 0.7053 0.2348 1.1467 1.9143 None 
STP-M002-00(114)01 50 6 0.4072 0.4110 0.1240 -0.0500 None 
STP-M002-00(116)01 51 9 0.5617 1.1861 0.7471 0.8111 None 
STP-M002-00(154)01 62 9 0.8029 0.4453 -0.3065 -0.2556 None 
STP-M002-00(155)01 44 7 0.7935 4.5029 0.3000 -0.3571 Variances 
STP-M002-00(165)01 118 10 2.5686 2.3129 -1.2975 -2.1800 None 
STP-M002-00(169)01 32 6 1.7600 1.3747 -0.0750 0.5333 None 
STP-M002-00(175)01 122 18 1.0972 1.0406 -0.2893 -0.6944 None 
STPN-12-1(110)01 44 20 4.2579 4.9531 -1.9705 -2.1450 None 
STP-M002-00(395)01 47 11 1.0893 1.2420 -0.8979 -1.0000 None 
STP-M002-00(236)01 32 7 0.4322 0.3495 0.5531 0.4429 None 
STP-M002-00(397)01 93 16 0.0000 0.0000 0.0000 0.0000 None 
STP-M001-00(330)01 87 10 2.0204 1.7250 -1.2460 -1.5500 None 
STP-M002-00(115)01 73 10 2.9465 4.1610 -1.6658 -1.9100 None 
 
 86
Table C-2.  ?? Project by Project Table for Comparison Tests 
Project # Nqct Ndot Vqct Vdot Mqct Mdot Differences
CM-186-1(28)01 7 7 16.8528 28.6191 -1.1429 -2.4286 None 
EDS-27-(136)01 8 8 1.9498 2.2755 -0.4875 -0.5125 None
EDS-27(147)01 9 9 7.1028 14.4894 -0.4444 0.3778 None 
EDS-84(16)01 7 7 11.4400 7.5295 1.5000 2.2429 None
EDS-84(17)01 11 11 7.0185 5.2827 2.2364 2.2455 None 
EDS-555(6)01 7 7 14.9491 10.1029 -0.1714 0.0571 None
GIP-341(34)01 14 14 18.8644 21.8710 2.3143 2.6214 None 
GIP-341(38)01 8 8 13.5041 16.6998 2.3875 3.2375 None
IM-NH-85-2(148)01 11 11 43.3587 39.1025 2.9545 4.0364 None 
NH-026-2(81)01 10 10 1.7271 3.6222 -0.5600 0.6000 None 
NH-56-1(62)01 8 8 10.6227 23.7070 1.9625 1.5125 None
NH-75-1(157)01 24 24 7.2089 8.4164 -0.2750 -0.2583 None 
NH-75-1(203)01 15 15 11.9492 16.7910 1.8067 1.3333 None 
NH-IM-75-1(158) 24 24 10.5087 8.1624 -0.1458 -0.0375 None 
NH-IM-95-1(119)01 7 7 3.2548 12.8995 -0.7143 0.2429 None 
NH-IM-95-1(125)01 15 15 3.8064 1.2511 -0.4933 -0.1400 None 
NH-IM-95-1(155)01 6 6 17.2667 7.4617 1.6667 4.6167 None 
NHS-M002-
00(398)01 26 26 7.0582 9.3768 0.2692 -0.3346 None 
NHS-M001-
00(530)01 20 20 14.3936 43.2898 2.8250 4.7350 None 
SAMA-39-(44)01 15 15 10.2470 7.2492 -0.3533 -0.9733 None 
SAMA-520(59)01 12 12 0.1715 0.7936 -0.2333 -0.5417 Variances 
STP-001-5(60)01 8 8 1.1543 0.3257 0.7000 0.3500 None 
STP-34-1(22)01 11 11 1.8096 1.2367 -2.0182 -1.1455 None 
STP-36-1(13)01 13 13 15.6133 9.4841 -1.2000 -1.6077 None 
STP-803(4)01 7 7 3.5462 6.1524 -0.8429 -0.0286 None
STP-M001-
00(273)01 15 15 3.1084 1.8927 -0.6133 -0.4867 None 
STP-M001-
00(330)01 11 11 0.6900 2.0980 -1.5000 -1.0000 None 
STP-M001-
00(773)01 8 8 2.5021 4.4541 0.2750 0.5625 None
STP-M002-
00(093)01 7 7 0.8357 1.0695 1.3714 1.3571 None 
STP-M002-
00(165)01 8 8 1.1841 8.8470 -1.0875 -0.6875 Variances 
STP-M002-
00(175)01 15 15 0.7812 1.4784 -0.2533 -0.0867 None 
STP-M002-
00(397)01 11 11 0.0000 0.0000 0.0000 0.0000 None 
STP-M001-
00(330)01 15 15 0.6554 1.7795 -1.6400 -1.1667 None 
STP-M002-
00(115)01 7 7 2.1824 0.5614 -1.8714 -1.2857 None
 
 87
Table C-3.  #200 Sieve Project by Project Table for QCT versus QA. 
Project # QC QA Vqc Vqa Mqc Mqa Differences
BRN-42-2(40)01 15 6 0.0392 0.1787 0.9267 0.7333 None 
CM-186-1(28)01 38 7 0.6502 0.1090 0.1895 -0.8286 Means 
CSSTP-M002-
00(444)01 40 9 0.5246 0.8700 -0.1050 0.2667 None 
CSSTP-M002-
00(453)01 53 7 0.4101 0.9567 0.2792 0.4000 None
EDS-27(136)01 140 35 0.3948 8.5014 1.0650 1.6086 None 
EDS-27(147)01 129 20 0.4037 0.3521 0.5605 0.6950 None
EDS-84(16)01 129 16 0.4262 0.8333 0.0295 -0.3000 None 
EDS-84(17)01 63 11 0.3243 0.3380 -0.1222 0.0000 None
EDS-545(8)01 72 9 0.1563 0.1011 -0.6583 -0.6111 None 
EDS-555(6)01 73 15 0.3415 0.5064 -0.6932 -0.4067 None
EDS-565(7)-01 99 7 0.0871 0.5981 -0.3798 -0.1143 Variances 
EDS-565-(15)01 98 15 0.1185 0.4054 -0.3622 -0.0400 Variances 
FLF-540(8)01 61 8 0.5134 0.4850 0.3620 0.5750 None 
FLF-540(25)01 92 10 0.2787 0.3129 0.1435 0.1200 None
FLF-540(28)01 116 10 0.2792 0.3877 -0.5802 -0.4100 None 
GIP-341(33)01 105 13 0.2229 0.2017 0.6114 0.4000 None
GIP-341(34)01 141 25 0.8528 0.5892 0.6624 0.4560 None 
GIP-341(38)01 199 21 1.1388 0.6096 0.5005 0.7190 None
HPP-3717-00(300)01 36 16 0.3483 6.6850 -0.2028 0.6313 Variances 
HPP-STP-178-1(33)01 23 7 0.3036 0.1429 0.9783 1.4429 None 
IM-00MS(268)01 43 14 0.5347 0.4699 0.2047 0.3071 None
IM-285-1(349)01 29 8 0.3747 0.9457 0.6552 0.1500 None 
IM-0000-00(471)01 60 15 0.4679 0.4724 0.4300 0.2667 None 
IM-0000-00(472)01 98 16 0.6396 0.8358 0.1357 -0.1625 None 
IM-00MS(20)01 56 17 0.4090 1.2307 0.5607 0.6947 Variances 
IM-MS(325)01 31 7 0.4046 0.4162 0.8452 1.0429 None 
IM-NH-85-2(148)01 175 20 0.6567 0.6519 0.3851 0.2350 None 
LAR00-S005-
00(459)01 21 7 0.5155 0.4762 -0.4381 0.2429 None
LAR00-S005-
00(593)C1 13 8 0.5286 1.2514 1.6769 1.0000 None 
LAR00-S005-00(636) 21 9 0.1613 0.0761 1.0857 1.1111 None 
LAR00-S005-
00(649)C1 13 6 0.2881 0.4857 1.2154 1.2167 None
LAR00-S005-
00(569)C1 18 7 1.1331 0.1924 -0.3611 -0.5714 None 
LAR31-87-1(067)01 26 6 0.2552 3.9017 0.6654 -0.8833 Variances 
LAR32-2-3(067)01 17 6 0.4088 0.3057 0.2588 0.3167 None 
LAR32-2-5(63)C1 12 6 0.1663 0.5027 0.6083 0.1333 None 
LAR32-24-1(057)C1 43 8 0.4962 2.0813 0.4953 0.4125 Variances 
LAR32-30-2(089)01 34 9 0.7408 1.5775 1.1088 0.1667 None 
LAR32-85-1(77)C1 20 6 0.1887 0.4560 -0.5650 0.1000 Means 
LAR32-250-1(057)C1 37 8 0.2692 0.7364 0.1270 -0.0250 None 
LAR32-888-1(089)01 45 13 0.6898 0.9869 0.3556 -0.0769 None 
 88
LAU32-8532-79(121)1 28 10 0.8504 0.7499 0.5179 0.0900 None 
NH-11-1(44)01 34 11 0.8803 0.4045 1.0294 0.2364 None
NH-012-1(83)01 59 10 0.2298 0.3529 1.2051 1.0200 None 
NH-017-2(53)01 36 13 0.7277 1.2076 0.6167 0.5615 None
NH-20-2(183)01 240 35 0.6932 1.2740 0.0221 -0.1686 Variances 
NH-026-2(81)01 29 9 0.8698 0.1525 0.8138 0.8000 Variances 
NH-043-1(49)01 15 6 0.3600 1.1800 0.3000 -0.1000 None 
NH-56-1(62)01 53 17 0.2916 0.9456 0.1811 0.0059 Variances 
NH-75-1(157)01 234 30 1.1829 1.5977 0.3821 0.6467 None 
NH-75-1(203)01 178 18 0.4697 0.5379 0.4685 0.4556 None
NH-171-1(4)01 100 8 0.7669 1.0584 0.4790 -0.0875 None 
NH-8013(8)01 64 20 0.2961 0.3768 0.4141 0.2000 None
NH-IM-75-1(158) 174 24 0.1503 0.1039 -0.4598 -0.7708 Means 
NH-IM-95-1(119)01 94 14 0.3517 0.7376 -0.2553 -0.1286 None 
NH-IM-95-1(125)01 142 29 0.4681 0.6280 0.5923 0.6962 None 
NH-IM-95-1(155)-01 117 12 0.4071 2.0906 0.2256 -1.2833 Both 
NHS-0000-00(768)01 60 8 0.2364 0.3155 -0.5517 0.2875 Means 
NHS-M002-00(398)01 487 53 0.8156 1.4086 0.3357 0.1094 Variances 
NHS-M000-00(436)01 26 6 0.7331 1.5347 0.4115 0.5667 None 
NHS-M001-00(530)01 193 10 1.5599 0.9423 0.5508 0.2700 None 
NHS-M001-00(531)01 59 12 0.1680 0.4027 0.0915 0.1500 None 
NHS-M001-00(591)01 50 13 0.8509 1.0647 -0.3640 -0.4846 None 
NHS-M002-00(276)01 25 8 0.5574 0.4498 0.3920 0.3125 None 
NHS-MOO2-
00(292)01 248 14 0.6024 0.5181 0.6867 0.8500 None 
PR000-S005-
00(810)C1 18 7 0.2965 0.5014 1.2333 0.1857 Means 
PR-147-2(055)01 13 9 0.7323 1.2100 1.5692 1.5000 None 
PR-3-3(151)01 20 6 0.4992 0.4057 -0.0350 0.2167 None
PR131-1(63)C1 8 8 0.1193 0.4279 0.3250 -0.5250 Means 
PR-1963-2(139)01 28 6 0.3417 0.1350 -0.6393 -0.3500 None 
PR-8530-83(39)C1 22 7 0.2492 0.5514 0.9818 0.0857 Means 
PR-8531-21(77)C1 50 10 0.3372 0.3107 0.3420 0.9800 Means 
PRLOP-8530-
32(15)C1 33 11 0.4761 0.8736 1.1212 0.9818 None 
SAMA-3(249)01 39 9 0.2983 5.2725 0.6897 -0.7333 Variances 
SAMA-3(294)01 123 25 0.2046 0.4732 0.6016 0.2640 Variances 
SAMA-3(298)01 238 31 0.6537 0.9971 1.0693 0.6226 Means 
SAMA-39(44)01 70 18 0.4312 0.4536 0.5314 0.4222 None 
SAMA-53(133)01 34 6 0.2912 0.1257 1.1971 0.8833 None
SAMA-60(5)01 51 10 1.8202 1.6254 0.0020 -0.3100 None 
SAMA-74(56)01 26 9 0.4256 0.8944 0.0000 -0.5222 None
SAMA-74(58)01 121 27 0.3228 0.2156 0.3521 0.4556 None 
SAMA91-(44)01 17 7 0.2328 0.1262 0.3176 0.4571 None
SAMA-155(52)01 18 6 0.3272 0.1080 -0.2389 0.2000 None 
SAMA-206-CO(3)01 71 14 0.8426 1.4037 0.2169 0.1714 None 
SAMA-234-(17)01 51 8 0.2353 0.1964 0.2471 0.0250 None 
SAMA-520(58)01 84 18 0.3688 0.4296 0.8512 0.9389 None
 89
SAMA-520(59)01 144 21 0.4100 0.7153 0.9639 0.6333 None 
SAM-M002-00(405)01 40 6 0.4567 0.3870 0.0350 0.3500 None 
SAM-M002-00(417)01 46 7 0.5425 0.7162 0.5130 -0.0429 None 
SAM-M002-00(422)01 64 17 0.3435 0.1803 1.1063 0.7176 None 
SAM-M002-00(399)01 23 7 0.3598 1.1314 0.8565 0.5143 None 
STP-001-5(60)01 37 9 0.0835 0.1050 0.6378 0.7333 None
STP-9-2(78)01 29 12 0.2969 0.4627 1.1759 1.2500 None 
STP-9-2(90)01 22 8 0.7004 0.9312 0.9682 0.5375 None
STP-34-1(22)01 89 14 0.3345 0.6146 -0.2809 -0.1071 None 
STP-36-1(13)01 52 15 0.3473 0.5112 1.3827 1.1133 None
STP-042-2(49)01 81 8 0.4149 0.5684 0.7741 0.7375 None 
STP-141-1(7)01 36 21 0.2936 1.6669 0.7111 -0.5905 Both 
STP-149-1(29)01 62 8 0.1694 1.1993 0.4677 -0.0250 Variances 
STP-803(4)01 72 18 0.4468 0.1697 1.0861 1.3556 None 
STP-2434(2)01 11 6 0.0720 1.8657 -0.1000 0.1167 Variances 
STP-4067(2)02 96 7 0.2505 0.1562 0.6521 0.6429 None 
STP-5121(3)01 34 7 0.2820 0.5433 0.5912 0.6000 None
STP-M00-00(703)01 54 9 0.8410 1.3644 -0.1593 -0.0222 None 
STP-M001-00(273)01 102 18 0.3875 0.4352 0.5863 0.4889 None 
STP-M001-00(490)01 24 6 0.7910 0.3657 0.8167 0.5833 None 
STP-M001-00(492)01 54 13 0.4932 0.1023 0.0130 0.2692 Variances 
STP-M001-00(493)01 121 13 0.3546 0.3967 -0.9694 -0.7000 None 
STP-M001-00(773)01 76 9 0.4820 0.6175 0.5737 0.4667 None 
STP-M001-00(872)01 77 6 0.3587 0.4587 0.2390 0.2667 None 
STP-M002-00(92)01 44 8 1.0007 1.4827 -0.0932 -0.3625 None 
STP-M002-00(093)01 45 7 0.2751 0.2048 0.8889 0.6143 None 
STP-M002-00(114)01 60 6 0.2947 0.0707 0.2083 -0.0333 None 
STP-M002-00(116)01 51 11 0.1184 0.0756 -0.4608 -0.0818 Means 
STP-M002-00(154)01 80 13 0.5608 0.4933 0.7650 0.5000 None 
STP-M002-00(155)01 44 7 0.3935 0.4514 0.2000 0.3857 None 
STP-M002-00(165)01 118 10 0.3073 0.5934 0.2093 -0.3300 Means 
STP-M002-00(169)01 32 6 0.4019 0.4120 1.0250 1.1000 None 
STP-M002-00(173)01 20 7 0.2333 0.2124 1.1800 0.7286 None 
STP-MOO2-00(175)01 122 18 0.1545 0.1375 0.5992 0.5889 None 
STP-M002-00(232)01 33 12 1.0950 1.8075 0.3242 -0.2750 None 
STP-M002-00(236)01 32 7 0.3646 0.2790 1.0156 1.1286 None 
STP-M002-00(397)01 93 16 0.5557 3.2172 0.6430 0.6625 None 
STP-M001-00(330)01 88 24 0.2392 0.4448 0.1057 -0.2042 None 
STP-M002-00(115)01 73 10 0.1716 0.1246 -0.0849 -0.0700 None 
STP-M002-00(395)01 68 13 0.7418 0.4100 0.3103 0.3000 None 
STPN-12-1(110)01 45 20 0.5239 0.7803 0.8200 0.6850 None 
 
 90
Table C-4.  #200 Sieve Project by Project Table for Comparison Tests. 
Project # Nqc Ndot Vqc Vdot Mqc Mdot Differences
CM-186-1(28)01 7 7 1.1767 0.4124 -0.2000 -0.8714 None 
CSSTP-M002-
00(444)01 6 6 0.4907 1.1177 -0.2667 -0.1167 None 
EDS-27(136)01 6 6 0.1830 0.2840 0.9500 0.9000 None
EDS-27(147)01 16 16 0.2543 0.3730 0.3313 -0.1312 None 
EDS-84(16)01 6 6 0.0987 0.2297 -0.2333 -0.1167 None
EDS-84(17)01 11 11 0.3342 0.4327 -0.1273 -0.2545 None 
EDS-555(6)01 7 7 0.1329 0.1614 -0.9571 -0.7143 Means 
GIP-341(34)01 14 14 0.5305 0.6394 0.7857 0.6357 None 
GIP-341(38)01 8 8 0.9164 0.7971 1.0750 1.1000 None
IM-MS(325)01 6 6 0.1870 0.6058 1.0500 1.2250 None 
IM-NH-85-2(148)01 11 11 1.1565 3.5836 1.0636 1.3182 None 
NH-20-2(183)01 7 7 0.5357 0.8162 -0.5286 -0.4429 None
NH-026-2(81)01 10 10 0.8588 1.5329 0.7900 0.4200 None 
NH-56-1(62)01 8 8 0.2427 0.6784 0.0375 0.0875 None
NH-75-1(157)01 22 22 0.8371 0.8606 0.5909 0.4591 None 
NH-75-1(203)01 15 15 0.5270 0.5735 0.6133 0.3933 None
NH-IM-75-1(158) 24 24 0.2786 0.3409 -0.4067 -0.6000 None 
NH-IM-95-1(119)01 7 7 0.1029 1.0857 -0.2571 0.6286 Variability 
NH-IM-95-1(125)01 15 15 0.5464 1.9098 0.6933 1.0533 None 
NH-IM-95-1(155)01 10 10 0.3600 2.1329 0.1000 -0.7800 Variability 
NHS-M002-00(398)01 25 25 0.5207 1.2893 0.5640 0.3480 None 
NHS-M001-00(530)01 19 19 1.3956 5.1427 0.1000 0.3053 None 
NHS-M002-00(294)01 12 12 0.3481 0.9661 0.8917 -0.1333 Means 
PR-128-2(065)01 6 6 0.4720 0.3257 0.9000 0.9167 None 
SAMA-39(44)01 15 15 0.4210 0.3070 0.6333 0.4467 None
SAMA-45(37)01 6 6 0.1977 0.6817 1.3167 2.0833 None 
SAMA-520(59)01 12 12 0.2499 0.5130 0.7917 1.0250 None
STP-001-5(60)01 8 8 0.1827 0.1400 0.5625 0.1000 None 
STP-9-2(78)01 6 6 1.7617 3.8827 0.7167 -0.1333 None
STP-34-1(22)01 11 11 0.4016 0.5000 -0.3182 -0.4000 None 
STP-36-1(13)01 13 13 1.5191 0.6547 1.0077 1.3154 None
STP-803(4)01 7 7 0.2533 1.1114 1.2000 1.0143 None 
STP-911(7)01 6 6 0.8987 0.1817 0.3333 0.3833 None
STP-M00-00(703)01 7 7 1.1781 1.1733 0.4143 -0.4000 None 
STP-M001-00(273)01 15 15 0.5150 0.2946 0.5267 0.7200 None 
STP-M001-00(773)01 7 7 0.4233 0.4029 1.0000 0.6571 None 
STP-M002-00(093)01 7 7 0.2562 0.5362 0.9571 0.7571 None 
STP-M002-00(114)01 6 6 0.1787 0.2937 -0.0667 -0.3167 None 
STP-M002-00(165)01 8 8 0.1927 0.2612 0.3125 0.0875 None 
STP-M002-00(175)01 15 15 0.1012 0.6397 0.5533 0.2600 Variability 
STP-M002-00(397)01 11 11 0.2227 0.9502 0.5545 -0.0727 None 
STP-M001-00(330)01 11 11 0.1849 0.3567 0.0091 -0.2455 None 
STP-MOO2-00(115)01 9 9 0.0961 0.1400 -0.0889 -0.0333 None 
 91
STP-MOO2-00(252)01 6 6 6.0297 4.0067 -1.0167 -0.8333 None 
STP-MOO2-00(395)01 6 6 0.6257 1.5240 0.3167 0.6000 None 
 92
Table C-5.  Asphalt Content Project by Project Table for QCT versus QA. 
Project # QC QA Vqc Vqa Mqc Mqa Differences
BRN-42-2(40)01 13 6 0.0083 0.0168 0.0569 0.0483 None 
CM-186-1(28)01 38 6 0.0370 0.0204 0.0021 0.1550 None 
CSSTP-M002-00(444)01 36 9 0.0217 0.0421 0.0011 -0.1222 None 
CSSTP-M002-00(453)01 53 7 0.0159 0.0337 0.0225 -0.0457 None 
EDS-27(136)01 138 32 0.0159 0.0176 0.0258 0.0094 None
EDS-27(147)01 130 20 0.0102 0.0196 -0.0417 -0.0030 None 
EDS-545(8)01 72 9 0.0309 0.0183 0.0275 0.1522 None
EDS-555(6)01 73 15 0.0384 0.1352 0.0403 0.2260 Variances 
EDS-565-(15)01 98 15 0.0364 0.0301 0.0040 0.0000 None 
EDS-565-(7)01 99 7 0.0233 0.0330 -0.0420 -0.0343 None
EDS84(16)01 130 16 0.0211 0.0298 0.0192 -0.0888 Mean 
EDS-84(17)01 65 11 0.0159 0.1757 0.0809 0.1364 Variances 
FLF-540(25)01 95 13 0.0189 0.0494 0.0002 0.0315 Variances 
FLF-540(28)01 116 10 0.0317 0.0295 0.0391 0.1580 None 
FLF-540(8)01 61 8 0.0250 0.0135 0.0816 -0.0300 None
GIP-341(33)01 107 13 0.0248 0.0209 0.0183 -0.1062 Mean 
GIP-341(34)01 145 25 0.1082 0.1254 0.1427 0.0468 None 
GIP-341(38)01 203 21 0.0845 0.1516 -0.0118 -0.0029 None
HPP-3717-00(300)01 36 16 0.0286 0.0355 -0.0653 -0.0413 None 
HPP-STP-178-1(33)01 23 7 0.0363 0.0038 -0.2222 -0.1800 Variances 
IM-0000-00(471)01 57 15 0.0396 0.0498 0.0974 0.0660 None 
IM-0000-00(472)01 73 10 0.0443 0.0812 0.0164 0.0230 None 
IM-00MS(20)01 56 15 0.0546 0.0852 -0.0636 0.0220 None
IM-00MS(268)01 41 14 0.0231 0.0472 0.0463 0.0214 Mean 
IM-285-1(349)01 30 8 0.0331 0.1437 -0.0077 0.1963 Variances 
IM-MS(325)01 31 7 0.0200 0.0115 -0.0710 0.0757 None 
IM-NH-85-2(148)01 175 20 0.0674 0.1105 -0.0145 0.0080 None 
LAR00-S005-
00(459)01 21 7 0.0336 0.2947 -0.0795 -0.0186 Variances 
LAR00-S005-
00(593)C1 13 8 0.0689 0.0667 -0.0985 -0.1687 None 
LAR00-S005-
00(649)C1 13 6 0.0728 0.0382 -0.0062 -0.4367 Mean 
LAR00-S005-
00(666)C1 8 7 0.0532 0.0372 0.0825 -0.0200 None 
LAR31-87-1(067)01 25 7 0.0181 0.0485 -0.0904 -0.0157 None 
LAR32-2-3(067)01 18 6 0.0357 0.0412 -0.0661 0.0333 None 
LAR32-888-1(089)01 44 12 0.0583 0.0312 -0.0925 -0.0917 None 
LAU32-8532-
79(121)01 16 6 0.0339 0.1333 0.2031 0.2667 None
NH-012-1(83)01 60 10 0.0353 0.0256 -0.1027 -0.1170 None 
NH-017-2(53)01 36 13 0.0464 0.0726 -0.0711 0.0915 None
NH-026-2(81)01 29 9 0.0420 0.0417 -0.0300 -0.0456 None 
NH-043-1(49)01 15 6 0.0462 0.0339 0.1200 0.0232 None
NH-11-1(44)01 34 10 0.0732 0.0590 0.0921 -0.0550 None 
 93
NH-171-1(4)01 100 8 0.0448 0.0351 -0.0253 -0.0500 None 
NH-20-2(183)01 240 29 0.0326 0.0519 -0.0117 0.0179 None
NH-56-1(62)01 53 16 0.0787 0.1061 0.0721 -0.1400 None 
NH-75-1(157)01 234 29 0.0584 0.0438 0.0233 0.0093 None
NH-75-1(203)01 180 18 0.0223 0.0314 -0.0245 0.0022 None 
NH-8013(8)01 64 15 0.0385 0.0303 0.0587 0.0353 None
NH-IM-75-1(158) 174 21 0.0236 0.0186 -0.0389 -0.0495 None 
NH-IM-95-1(119)01 90 14 0.0303 0.0449 0.0006 -0.0493 None 
NH-IM-95-1(125)01 142 29 0.0210 0.0223 -0.0182 -0.0152 None 
NH-IM-95-1(155)-01 117 12 0.0538 0.0561 0.0549 0.0883 None 
NHS-0000-00(768)01 61 8 0.0330 0.0672 0.0631 0.0763 None 
NHS-M001-00(530)01 193 9 0.0499 0.0667 -0.0331 -0.0756 None 
NHS-M001-00(531)01 59 12 0.0291 0.0241 -0.0478 -0.0733 None 
NHS-M001-00(591)01 51 13 0.0554 0.0584 -0.0816 -0.0469 None 
NHS-M002-00(276)01 25 8 0.0224 0.0258 -0.0156 0.1525 None 
NHS-M002-00(292)01 251 10 0.0425 0.0453 0.0615 0.0040 None 
NHS-M002-00(398)01 487 53 0.0388 0.0613 -0.0556 -0.1036 Variances 
PR-147-2(055)01 13 9 0.0698 0.1396 0.0223 -0.0133 None 
PR-1963-2(139)01 28 6 0.0678 0.0768 0.0239 -0.0400 None 
PR-3-3(151)01 21 6 0.0521 0.0305 0.0614 -0.0267 None
PR-8530-83(39)C1 22 10 0.0327 0.0500 -0.0873 0.0960 None 
PR-8531-21(77)CT1 49 10 0.0482 0.0981 -0.0059 0.0810 None 
PRLOP-8530-
32(15)C1 31 11 0.0375 0.0339 0.0587 -0.0236 None
SAMA-155(52)01 18 6 0.0134 0.0294 -0.0983 -0.1617 None 
SAMA-206-C0(3)01 70 14 0.0504 0.0327 0.0271 0.1279 None 
SAMA-234-(17)01 51 8 0.0111 0.0058 -0.0553 -0.0500 None 
SAMA-3(249)01 38 9 0.0264 0.1858 0.0618 0.1167 Variances 
SAMA-3(294)01 123 25 0.0123 0.0551 0.1577 0.1488 Variances 
SAMA-3(298)01 229 31 0.0579 0.0740 0.0089 -0.0258 None 
SAMA-39(44)01 70 18 0.0393 0.0218 -0.0560 -0.1294 None
SAMA-520(58)02 84 18 0.0255 0.0280 0.0692 -0.0117 None 
SAMA-520(59)01 144 21 0.0405 0.0416 0.1119 0.0352 None 
SAMA-53(133)01 34 6 0.0175 0.0062 -0.0729 -0.0350 None
SAMA-60(5)01 50 10 0.0546 0.0628 -0.0544 -0.0580 None 
SAMA-74(56)01 25 9 0.0346 0.0646 0.0312 0.0911 None
SAMA-74(58)01 121 27 0.0402 0.0089 -0.0235 -0.0867 Variances 
SAM-M002-00(399)01 23 7 0.0381 0.0922 -0.0026 0.1200 None 
SAM-M002-00(405)01 40 6 0.0202 0.0426 0.0015 -0.0550 None 
SAM-M002-00(417)01 46 7 0.0209 0.0386 0.0654 -0.1143 Mean 
SAM-M002-00(422)01 64 17 0.0261 0.1004 0.0548 0.1988 Variances 
STP-001-5(60)01 37 9 0.0367 0.0196 -0.0895 0.0267 None 
STP-042-2(49)01 82 8 0.0386 0.0438 0.1521 0.0100 None
STP-141-1(7)-01 36 21 0.0274 0.0346 0.0008 0.0524 None 
STP-149-1(29)01 62 8 0.0454 0.1082 0.0218 -0.1037 None
STP-2434(2)01 14 6 0.0161 0.0208 0.0814 0.0867 None 
STP-34-1(22)01 89 14 0.0159 0.0192 0.0201 0.1157 None
STP-36-1(13)01 49 15 0.0531 0.0564 -0.0463 -0.0660 None 
 94
STP-4067(2)02 98 7 0.0275 0.0251 0.0756 0.1243 None 
STP-5121(3)01 34 7 0.0088 0.0094 0.1921 0.2071 None
STP-803(4)01 73 17 0.0222 0.0238 -0.0025 0.0135 None 
STP-9-2(78)01 29 12 0.0120 0.0221 -0.0293 0.0258 None
STP-9-2(90)01 22 8 0.0221 0.0205 0.0732 0.1100 None 
STP-M00-00(703)01 53 9 0.0208 0.0257 0.0740 -0.0056 None 
STP-M001-00(273)01 100 18 0.0298 0.0227 0.0087 -0.1083 Mean 
STP-M001-00(330)01 88 14 0.0437 0.0318 0.0102 -0.0286 None 
STP-M001-00(490)01 25 6 0.1374 0.1993 0.1120 0.1667 None 
STP-M001-00(492)01 54 13 0.0125 0.0160 -0.0256 -0.0508 None 
STP-M001-00(493)01 121 13 0.0185 0.0315 -0.0264 -0.0169 None 
STP-M001-00(773)01 76 9 0.0175 0.0438 0.0080 0.0344 None 
STP-M001-00(872)01 77 7 0.0164 0.0225 0.0522 0.1257 None 
STP-M002-00(093)01 45 7 0.0075 0.0206 0.0602 0.1200 None 
STP-M002-00(114)01 60 6 0.0109 0.0267 0.0560 0.0583 None 
STP-M002-00(115)01 73 10 0.0088 0.0155 -0.0026 -0.0070 None 
STP-M002-00(116)01 51 11 0.0111 0.0109 0.0408 0.0409 None 
STP-M002-00(154)01 80 13 0.1161 0.3721 0.0769 0.0223 Variances 
STP-M002-00(155)01 44 7 0.0212 0.0396 0.0116 -0.1457 None 
STP-M002-00(165)01 118 10 0.0089 0.0142 0.0168 -0.0350 None 
STP-M002-00(169)01 32 6 0.0157 0.0241 -0.1112 0.0583 Mean 
STP-M002-00(173)01 20 7 0.0223 0.0559 0.0745 0.0514 None 
STP-M002-00(232)01 33 12 0.0652 0.0598 -0.0370 0.0317 None 
STP-M002-00(236)01 32 7 0.0315 0.0210 0.1131 0.1000 None 
STP-M002-00(395)01 68 13 0.0414 0.0199 -0.0060 0.1046 None 
STP-M002-00(397)01 93 15 0.0240 0.0230 -0.0260 0.1087 Mean 
STP-M002-00(92)01 44 8 0.0315 0.0970 -0.1477 -0.0813 None 
 
 95
Table C-6.  Asphalt Content Project by Project Table for Comparison Tests. 
Project # Nqc Ndot Vqc Vdot Mqc Mdot Differences
CM-186-1(28)01 7 7 0.0518 0.0861 -0.0886 -0.0371 None 
CSSTP-M002-00(444)01 6 6 0.0218 0.0603 -0.0983 -0.3000 None 
EDS-27-(136)01 8 8 0.0231 0.0600 0.0038 -0.0062 None
EDS-27(147)01 16 16 0.0149 0.0345 0.0044 0.0762 None 
EDS-84(16)01 7 7 0.0344 0.0370 0.0371 0.0071 None
EDS-84(17)01 11 11 0.0201 0.0085 0.1382 0.0918 None 
EDS-555(6)01 7 7 0.1698 0.2422 0.1357 0.1857 None
GIP-341(34)01 14 14 0.1124 0.1143 0.2129 0.1757 None 
GIP-341(38)01 8 8 0.0593 0.1549 0.1225 0.1350 None
IM-MS(325)01 6 6 0.0153 0.0566 -0.0467 -0.2750 None 
IM-NH-85-2(148)01 11 11 0.2703 0.3765 0.1336 0.2991 None 
NH-20-2(183)01 7 7 0.0534 0.0363 -0.0257 -0.0757 None
NH-56-1(62)01 8 8 0.0946 0.1336 -0.0113 0.0025 None 
NH-75-1(157)01 24 24 0.0362 0.0942 -0.0092 0.0250 None
NH-75-1(203)01 18 18 0.0230 0.0299 -0.0183 0.0128 None 
NH-IM-75-1(158) 24 24 0.0455 0.0591 -0.1304 -0.0933 None
NH-IM-95-1(119)01 7 7 0.0142 0.0561 -0.0486 0.0814 None 
NH-IM-95-1(125)01 15 15 0.0197 0.0663 0.0060 -0.0507 None 
NH-IM-95-1(155)-01 10 10 0.0447 0.0442 0.0650 0.1710 None 
NHS-M002-00(398)01 29 29 0.0374 0.0773 -0.1055 -0.1072 None 
NHS-M001-00(530)01 18 18 0.0713 0.2855 0.0811 0.1372 Variances 
PR-8530-83(39)C1 6 6 0.0489 0.0527 -0.0050 -0.0100 None 
SAMA-39-(44)01 15 15 0.0072 0.0262 -0.0447 -0.1393 None
SAMA-206-CO(3)01 9 9 0.0393 0.1971 0.1389 0.3544 None 
SAMA-520(59)01 12 12 0.0529 0.0514 -0.0025 0.0383 None
STP-037-2(60)01 11 11 0.1184 0.2392 -0.1364 -0.0755 None 
STP-9-2(78)01 6 6 0.0179 0.0190 -0.0683 -0.0600 None
STP-34-1(22)01 11 11 0.0214 0.0368 0.0055 0.0164 None 
STP-36-1(13)01 13 13 0.0525 0.1435 -0.0008 0.1223 None
STP-803(4)01 7 7 0.0305 0.0742 -0.0629 -0.1957 None 
STP-M00-00(703)01 7 7 0.0270 0.0287 0.0529 -0.0471 None 
STP-M001-00(273)01 15 15 0.0156 0.0344 -0.0167 -0.1027 None 
STP-M001-00(330)01 11 11 0.0643 0.0340 0.0691 0.0736 None 
STP-M001-00(773)01 8 8 0.0100 0.0377 0.0675 0.0750 None 
STP-M002-00(093)01 7 7 0.0199 0.0128 0.0571 0.0986 None 
STP-M002-00(114)01 6 6 0.0111 0.0928 0.0383 0.0000 None 
STP-M002-00[116]01 6 6 0.0063 0.0078 0.1000 0.0783 None 
STP-M002-00(165)01 8 8 0.0063 0.0093 0.0437 -0.0250 None 
STP-M002-00(175)01 15 15 0.0229 0.0440 0.0140 0.0920 None 
STP-M002-00(397)01 11 11 0.2125 0.2669 -0.1718 0.0227 Means 
STP-M002-00(115)01 7 7 0.0031 0.0200 -0.0257 -0.0357 None 
 96
APPENDIX D 
PROJECT BY PROJET ANALYSIS SCATTERPLOTS 
0
5
10
15
20
25
30
35
40
45
2040608010120
QA
QC
T
 
Figure D-1.  Project by Project Variances for ?? Sieve QCT versus QA. 
-3
-2
-1
0
1
2
3
-3 -2 -1 0 1 2 3
Average QA - JMF (% Passing)
A
ver
a
g
e Q
C
 
- 
JMF
 (%
 P
asg
g
i
n
g
)
Linear (Line of
Absolute Numerical
Equality)
 
Figure D-2.  Project by Project Average Means for ?? Sieve QCT versus QA. 
 97
0
5
10
15
20
25
30
35
40
45
50
0 5 10 15 20 25 30 35 40 45 50
DOT
QC
T
 
Figure D-3.  Project by Project Variances for ?? Sieve for Comparison Tests. 
-5
-4
-3
-2
-1
0
1
2
3
4
5
-5 -4 -3 -2 -1 0 1 2 3 4 5
Average DOT-JMF (% Passing)
A
v
e
r
ag
e Q
C
T
 -
 
J
M
F
 (
%
 P
assi
n
g
)
 
Figure D-4.  Project by Project Means for ?? Sieve for Comparison Tests. 
 98
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0123456789
QA
QC
 
Figure D-5.  Project by Project Variances for #200 Sieve QCT versus QA. 
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Average QA-JMF (% Passing)
A
v
e
r
ag
e Q
C
-JMF
 (
%
 
P
a
ssi
n
g
)
 
Figure D-6.  Project by Project Average Means for #200 Sieve QCT versus QA. 
 99
0
1
2
3
4
5
6
7
1234567
QA
QC
T
 
Figure D-7.  Project by Project Variances for #200 Sieve Comparison Tests. 
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5
Average DOT-JMF (% Passing)
A
v
e
r
a
g
e
 QC
T-
J
M
F (
%
 Pa
s
s
i
n
g)
 
Figure D-8.  Project by Project Average Means for #200 Sieve Comparison Tests. 
 100
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
QA
QC
T
 
Figure D-9.  Project by Project Variances for Asphalt Content QCT versus QA. 
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Average QA-JMF (% Passing)
A
v
er
a
g
e 
Q
C
-
J
MF
 (%
 Pa
s
s
in
g
)
 
Figure D-10.  Project by Project Average Means for AC for QCT versus QA. 
 101
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
DOT
QC
T
 
Figure D-11.  Project by Project Variances for AC for Comparison Tests. 
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
Average DOT-JMF (Asphalt Content)
Av
e
r
ag
e
 Q
C
T
 (As
p
h
a
l
t C
o
n
t
en
t)
 
Figure D-12.  Project by Project Average Means for AC for Comparison Tests.