Model selection in orthogonal regression

Abstract

We develop the relationship between the stepwise F-test model selection criteria and information-based criteria for orthogonal regression models. We obtain the asymptotic properties of the stepwise F-tests with respect to efficiency and consistency. The performances of F-test as well as other efficient and consistent criteria are compared via a large scale simulation study. The results indicate that three of the F-test criteria should be considered for routine data analysis.