Oracle Inference for Misspecified Models

• Activity Number: 168
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: IMS
• Author(s): Mihai Giurcanu*+ and Brett Presnell
• Companies: University of Florida and University of Florida
• Keywords: misspecified models ; spatio-temporal processes ; block-bootstrap ; moment condition models ; oracle property ; efficient estimators

Abstract

We have developed a new class of estimators for misspecified moment condition models for stationary time series. These estimators have an oracle property: their large sample behavior is the same as of efficient estimators obtained under the apriori information that the true moment conditions were known in advance. We present asymptotic inferential procedures for these estimators as well as of various block-bootstrap estimators of their distributions and of null distributions of relevant test statistics. The results of an extensive simulation study show the finite sample properties of the asymptotic and block-bootstrap oracle inferential procedures for a stationary dynamic spatial process with misspecified random effects. This research is joint with Dr Brett Presnell.