58
reads

Wei-Yann Tsai

2011-05-20
15:00:00 - 16:30:00

307 , Mathematics Research Center Building (ori. New Math. Bldg.)

We propose a semiparametric proportional likelihood ratio model which relates the covariates and the baseline density function. This model generalizes Gilbert-Lele-Vardi's selection bias model by allowing the weight function to depend on both covariates and outcomes, and extends the generalized linear models by leaving the distribution unspecified. Maximum likelihood estimator and moment-type estimators are proposed and their asymptotic properties are derived. A small simulation study shows that the proposed estimators perform well compared with the gold-standard parametric maximum likelihood estimator.