Simultaneous Modeling of Multiple Outcomes Over Time: A Hierarchical Modeling Approach
Das, A., & Gard, C. C. (2004, August). Simultaneous Modeling of Multiple Outcomes Over Time: A Hierarchical Modeling Approach. Presented at , .
Multiple outcomes assessed longitudinally over time are common inpublic health research. This work, motivated by the Maternal LifestyleStudy, involves children prenatally exposed to cocaine prospectivelyfollowed up for the onset of abnormal behavior problems assessed byseveral standardized instruments measuring different domains ofbehavior. Thus, at each time point, each child contributed to multiple,possibly correlated, outcomes that collectively constitute onesyndrome, while specific outcomes affected by cocaine are also ofscientific interest. Since it is neither substantively appropriate norstatistically efficient to fit separate longitudinal models for eachoutcome, here we utilize the covariance structure in hierarchicalmodels to simultaneously fit all the behavioral outcomes as a functionof cocaine exposure and other covariates. The advantages of thisapproach are (a) it recognizes that all outcomes together constituteone syndrome, (b) it boosts statistical efficiency by incorporatingcorrelated outcomes and avoiding multiple comparisons, and (c) itallows for outcome-specific exposure effects so we can identify thespecific behavioral domains affected by cocaine.