Storck, S.; Kastner, C.; Toutenburg, Helge
Longitudinal data with dropouts: a comparison of pattern mixture models with complete case analysis.
Sonderforschungsbereich 386, Discussion Paper 188
Pattern mixture models constitute an alternative to selection models (Little & Rubin, 1987). Little & Wang (1996) introduced pattern mixture models for analyzing multivariate normal longitudinal data with missing values. This paper was the theoretical foundation and the induce to investigate the small sample properties of pattern mixture models compared with complete case analysis. The main point of interest, of the simulations, was the mean square error of the estimated model parameters. Parameters estimated by the pattern mixture model are very satisfying under ignorable mechanism but they have to be scanned carefully under nonignorable mechanism.