|
Abstract:
|
A matched case-crossover study is a blend of a matched case-control study and crossover design. The typical case-crossover design uses a sample from a study population of individuals, all of whom have experienced the outcome of interest. Measurements are taken on each subject in an exposed and unexposed setting; each subject acts as their own control. Conditional logistic regression is often used in case-crossover design. However, the analyses can be complicated in semiparametric functional analysis settings with many covariates for two reasons. First, not all of the covariates have known functional forms. Second, some functions are clustered to each other and the number of clusters is not always known in advance. Hence in this talk, we propose a flexible semiparametric functional approach which is developed using a weighted Dirichlet Process Mixture model (WDPM). An infinite mixture model with DP can solve the second problem easily because DP automatically clusters. Unknown functional forms are estimated via regression-splines and variable selections are conducted as well to reduce the high dimensionality. We show the advantage of our approach using matched case-crossover studies.
|
