Title: Semiparametric generalized linear models: Application to biased samples
Authors: Paul Rathouz - University of Texas at Austin (United States) [presenting]
Abstract: A novel class of generalized linear models indexed by a linear predictor and a link function for the mean of $(Y|X)$ has been previously proposed. In this class, the distribution of $(Y|X)$ is left unspecified and estimated from the data via exponential tilting of a reference distribution, yielding a response model that is a member of the natural exponential family. We focus on how, with very easy-to-implement modifications, the model can accommodate biased samples arising from two-phase extensions of case-control designs to count or continuous response distributions, wherein inferences about the mean are of interest.