Estimation procedures for mixed-effects models with applications to normally distributed and survival data

Abstract

In this dissertation we have considered several issues related to the estimation of mixed effects models. In Chapters 2 and 3 we have evaluated the performance of the simplified strategies, and also the impact of ignoring a level in fitting linear mixed effects models for normally distributed data. In Chapter 4 we have briefly illustrated some methodological issues related to the use of mixed effects models for failure-time data, as well as several estimation procedures to fit the models. In Chapter 5, we have proposed a novel modification of the EM algorithm, in which the Laplace approximation is used at the E-step. We have conducted a simulation study (Chapter 6) to evaluate the performance of the proposed estimation procedure and to compare it with alternative approaches. Finally, in Chapter 7 we have investigated the possibility of using simplified strategies in the meta-analytic approach to the validation of surrogate endpoints in the case of two failure-times.