Hua Liang, University of
Statistical Inference in Partially Linear Models with Missing Response Variables and Error-prone Covariates
We investigated partially linear models when the response variable is sometimes missing with missing probability depending on the covariates, and the linear covariate is measured with error. We proposed a class of semi-parametric estimators for parameter of interest. The resulting estimators were shown to be consistent and asymptotically normal under general assumptions. To construct a confidence region of the parameter and to avoid estimating covariance matrix because of its complexity, we also proposed an empirical likelihood based statistic, which was shown to have an asymptotic chi-squared distribution. Extensions to generalized partial linear models were also considered. The proposed methods were applied to analyze an AIDS clinical trial dataset. A simulation study was also reported to illustrate our approach.
Joint work with Drs. Suojin Wang and Raymond Carroll