|
Abstract:
|
In this paper, we propose a Bayesian semiparametric method to estimate heterogeneous direct and indirect effects of distinct clusters that are formed by the effect modifiers. These cluster-specific (or heterogeneous) direct and indirect effects can be estimated through a set of regression models whose individual-varying coefficients are clustered together by specifying a DP prior on the mixing distribution of the effect modifiers. Unlike other existing methods for the heterogenous effects (which require the pre-deteremed number of clusters), our model automatically detects/defines clusters. Also, to allow full flexibility of the models, we use a Bayesian semiparametric method to specify the outcome models.
|
