Joint modelling of multiple outcomes on longitudinal behaviour of genetically modified mice

Abstract

Aim of the study is to investigate the differences in the evolution over time between TauPS2APP and wild-type mice for the behaviour data. Moreover, to fit joint model combining outcomes in order to investigate if there is any benefit compared to a univariate analysis. Two datasets, extracted from a pre-clinical experiment on Alzheimer's disease, consisting information regarding 5 tests for different behaviours of young and old mice, are combined. Correlations among outcomes from same mouse are captured by modelling the responses jointly by fitting them pair-wise and combining the results using pseudo likelihood theory. Univariate generalized linear mixes models (GLMM) shows significant evidences of difference in the evolution over time between genotypes for all the behaviors of young mice as well as combined data. Multivariate joint model also shows the similar effect on overall responses. Effects of age and genotype are also significant.