Time for a change: Studying individual differences in dynamics
The aim of this project is to investigate the possibilities for studying intra- and inter-individual variability in intensive longitudinal data using Multilevel Latent Markov Models (LMMs). These models can be used with univariate or multivariate, and categorical, continuous or mixed data, which makes them especially useful for psychological studies. However, a proper investigation of inter-individual differences requires the inclusion of (multiple) random effects, and this may pose problems for the estimation of the model. The Bayesian framework seems to be quite robust but it remains to be seen which requirements should be met for feasible estimation. Simulations will be used to investigate the numbers of persons, measurements, and latent states that result in adequate estimates of the random and fixed effects. In addition, there are unsolved problems regarding model evaluation and comparison in the Bayesian framework, and how best to deal with label-switching, and how to deal with prior influence in the logit model underlying the latent state transitions. Simulations can be used to address these issues in more detail. Throughout the project, the models are also applied to observational data and questionnaire data, to illustrate the practical use and substantive interpretation of different types of LMMs.
NWO (Netherlands Foundation of Scientific Research)
Prof. dr. Herbert Hoijtink (Utrecht University)
Dr. Ellen Hamaker (Utrecht University)
1 September 2012 – 1 September 2016