Mixed-effects location scale models for ordinal data

Basic data for this project

Description

The extent to which individuals fluctuate in a variable over time plays a central role in a number of psychological theories. To investigate hypotheses regarding this intra-individual variability (also re-ferred to as within-person variability), researchers typically conduct an experience sampling or a daily diary study to collect repeated measures of the variable of interest. The resulting data is then analyzed using an extension of a longitudinal mixed-effects model called the mixed-effects location scale model (MELS). Currently, the MELS can only be used when the variable of interest is contin-uous. However, many variables in psychology have an ordinal scale, so that the MELS cannot be used for data analysis in these situations. The aim of the proposed project is to extend the MELS so that one or more ordinal outcome variables or a mix of ordinal and continuous outcome variables can be analyzed. For the estimation of the model parameters, a maximum likelihood approach will be derived and implemented in an R package. Furthermore, the performance of the proposed max-imum likelihood estimator will be compared with a Bayesian approach implemented in JAGS. Overall, we believe that the planned extensions to MELS will allow applied researchers to investi-gate a number of new and interesting research questions using appropriate statistical approaches.