|
Abstract:
|
Under the usual mixed model design, outcomes are often analyzed 'univariately.' In aging research, multiple systems are often studied simultaneously (e.g., ADLs, IADLs, cognitive impairment, physical performance, quality of life). Change over time or trajectories across these multiple systems are then correlated. Snijders (1999) and Thiebaut (2002) demonstrate a bivariate hierarchical model to simultaneously model 2 outcomes. The two level mixed model will be extended to see if, in longitudinal studies, (1) the 'trajectories' from the multiple outcomes are correlated, and (2) if classes of trajectories can be defined. Using data from the Duke EPESE study, change in measures of ADLs, IADLs, Cognitive Impairment, and Quality of Life will be estimated. Models correlating the empirical Bayes parameters of the trajectories estimated univariately will be contrasted to multivariate models.
|
