CMStatistics 2017: Start Registration
View Submission - CMStatistics
B0887
Title: A review of panel data models with a Markov dependent structure for univariate and multivariate ordinal responses Authors:  Fulvia Pennoni - University of Milano-Bicocca (Italy) [presenting]
Abstract: The purpose is to review the theory related to the Markov chain model and to the latent Markov model with univariate or multivariate ordinal responses. It frequently happens on social and medical surveys, and it is also common to encounter such data on other fields of applications by which the order can be made according to subject matter knowledge. We review how to assess the influence of covariates according to a chosen link function based on local, global, or continuation logits, or by an ordered probit link function. In this way, the model parameters can retain the ordinal structure imposed by the data. We show the statistical algorithms for the parameters estimation and through examples based on real data we show how predictions can be achieved.