A0340
Title: Estimation of dynamic Logit mixed models for multinomial responses with categorical covariates
Authors: Alwell Oyet - Memorial University (Canada) [presenting]
Brajendra Sutradhar - Memorial University (Canada)
Abstract: A situation is considered where multinomial responses are collected repeatedly over a short period of time from a large number of independent individuals along with individuals' categorical covariate information. Dynamic logit models are developed under the assumption that the responses are influenced by (a) The categorical covariates, (b) Past multinomial responses, and (c) Some category-prone unobservable variables. By category prone, we refer to unobservable variables that influence the responses into specific categories. The likelihood estimation of the effects of the covariates, the dynamic dependence parameters, and the variances of the category-prone random effects are discussed. The results of simulation studies in special cases of the longitudinal mixed model, namely, a cross-sectional model, longitudinal fixed effects model and a longitudinal mixed model, are discussed. Asymptotic properties of the covariate effects parameters may be discussed if time permits.
