Longitudinal methods for life course research : A comparison of sequence analysis , latent class growth models , and multistate event history models for studying partnership transitions

This paper compares and contrasts three methods that are useful for life course researchers; the more widely used sequence analysis, the promising but less often applied latent class growth models, and multistate event history models. The strengths and weaknesses of each method are highlighted by applying them to the same empirical problem. Using data from the Norwegian Generations and Gender Survey, changes in the partnership status of women born between 1955 and 1964 are modelled, with education as the primary covariate of interest. We show that latent class growth models and multistate event history models are a useful addition to life course researchers’ methodological toolkit and that these methods can address certain research questions better than the more commonly applied sequence analysis or simple event history analysis.