Abstract:
|
We propose and evaluate models for bivariate functional data collected longitudinally. A number of extensions is entertained including cross-outcome effects, models for non-commensurate outcomes (e.g. continuous and binary), variable domains and a hierarchical dependence structure. Specifically, effects of functional predictors on non-commensurate bivariate outcomes are estimated taking into account correlations between the outcomes induced by the shared random effects. We apply the proposed methods to the ecological momentary assessment data studying the dependence of the relationship satisfaction and condom use amongst young adults participating in an STD study.
|