Semiparametric Estimation for Integral Projection Models

Abstract

Understanding the connection between variation in climate and population dynamics of plants and animals is important for predicting the impacts of future climate change. A popular approach for studying population dynamics is integral projection models, in which covariates are easily parameterized by regression models. The main goal of this dissertation is to extend the scope of stochastic integral projection models (IPMs), which have been commonly fitted with linear or generalized linear models in the past.

The state-space models and the stochastic IPMs using linear mixed models are discussed in Chapter 2 and Chapter 3 respectively. In the state-space models, we study climate effects on the population, which is segmented in different age groups. We find that we may need to include climate effects to better understand the population dynamics. The second chapter provides an important basic analysis for the validity of the findings of the IPMs. The third chapter provides the analyses of the stochastic IPMs with linear and generalized linear models in which the LASSO method is applied for variable selection, and perturbation analyses are discussed.

In Chapter 4, by fitting more flexible IPMs, we develop a new method of finding elasticities of population growth rate to climate effects by estimating the derivatives of smooth functions of semi-parametric regression models.

Based on the models studied in this dissertation, we find that the projected population growth is consistent across all models. In addition, we find that climate variables associated with temperature have significant effects on the population growth rate.