A note on model diagnostics in longitudinal data analysis (Q2463668)

Due to the necessity of repeated measurements obtained from each subject most of the studies in biomedical research are longitudinal. This paper is focused on analyzing whether there exists any significant systematic (model) departure in the data from the model assumption or not. To remedy model diagnostics in longitudinal data analysis the authors propose eight testing procedures for randomness accompanied with some conventional and/or non-conventional plots and verify their applicability by independent data sets in SAS/STAT system. The results of the paper manifest that the peculiar diverseness of the dependence within subjects in longitudinal data can perplex the model diagnostics, and hence, show the necessity of using multiple indexes. The proposed testing procedures are illustrated by four medical studies in Taiwan.