Modelling Strategies for Repeated Multiple Response Data

Agresti and Liu (2001) discussed modelling strategies for a multiple response variable, a cat- egorical variable for which respondents can select any number of outcome categories. This article discusses modelling strategies of a repeated multiple response variable, a categorical variable for which respondents can select any number of categories on repeated occasions. We consider each of the responses as a binary response and model the mean binary responses with two approaches: a marginal model approach and a mixed model approach. For the marginal model approach, we consider a generalised estimating equations (GEE) method to account for dierent correlations over time and between items as an alternative to standard GEE, which only allow relatively simple correlation structures. We illustrate the dierent ap- proaches using The Household, Income and Labour Dynamics in Australia (HILDA) Survey, a household-based panel study.