时间:2018年4月17号上午11:00-12:00

地点:T7-705

题目:Basic Selections of Generalized Linear Models

主讲人:Professor Philip E. Cheng, Research Fellow of Institute of Statistical Science Academia Sinica, Taiwan

Abstract of the talk:
A generalized linear regression model is used to illustrate a response variable of interest in some experimental statistics of our daily life. Whatever large the data size and the variables dimension, the selection of a parsimonious model is the difficult goal to achieve. It is expected to choose minimal explanatory variables and minimal interaction terms altogether so that the most concise regression model is selected for use. In this study, we will introduce an elementary analysis of discrete multivariate histograms in which continuous variables are suitably discretized and multivariate multinomial likelihoods are employed to offer the desired main-effect variables and their interaction terms. Against standard model selection criteria, the proposed method is shown to offer the most concise model effects in a few empirical studies, before choosing the type of a regression model.