Subjects: Mathematics >> Statistics and Probability submitted time 2019-10-14
Abstract: Most authors have examined the inference of model averaging estima- tors under local misspecification assumption. [1] is one of the most groundbreaking and famous articles. However, although the local misspecification assumption pro- vides a suitable framework for studying the asymptotic theories of Frequentist Model Averaging estimators, it also draws comments from [2] because of its realism. In this paper, we prove that the confidence interval construction method for the linear function of parameters in [1] is still valid in linear regression with normally distributed error under general fixed parameter setup. It means that the coverage probability of the confidence interval can reach the nominal level. The simulation results support our theory conclusion. "
Peer Review Status:Awaiting Review