Nov. 4 2009 2:00PM

Nov. 4 2009 4:00PM

Generalized linear models have been introduced by Nelder and Wedderburn (1972) as a unifying family of models for nonstandard cross-sectional regression analysis with non-normal responses. We focus on multinomial response models that model relationships between a multinomial response variable and a set of explanatory variables. Based on phi-divergences an estimator of the generalized linear models for multinomial data (GLM) under linear restrictions on the parameters is considered. New test statistics, also based on phi-divergences are considered as alternatives to the classical ones for testing a hypothesis about linear restrictions on the parameters. The asymptotic distribution of them is obtained under the null hypothesis as well as under contiguous local hypotheses. A wide simulation study of the estimators and the test statistics is carried out. After fitting data by a certain generalized linear model, specific tools to select and check models are used. In this talk, a new selection criterion based on divergence measures is proposed in backward removal variable selection for multinomial generalized linear model. Finally, we extend some diagnostic tools using the phi-divergence measure for detecting outliers, leverage and influential points.

Pravidelný seminář Rozhodování a řízení za neurčitosti. (Contact person: Miroslav Kárný)