Title: Mixed models, Mixture models, Maximum likelihood and the letter "M"
Abstract: Correlated data is collected in almost every field of science, from gene expression studies in bioinformatics to community composition data in ecology. The analysis of correlated data remains an open question, since the statistical theory and methods we should apply depend critically on the nature of the correlation, the key properties of the data and the question of interest. In this seminar, I discuss some of the main branches of my current research, which revolves around the core theme of the statistics of correlated data. These include new methods and theory I have developed for model-based classification of correlated data in ecology, and for joint selection in mixed models applied to epidemiological studies. The seminar will combine more technical aspects with general interpretation and intuition behind the theory and methods proposed.