A misspecification test for beta prime regression models

Abstract

The beta prime regression model proposed by Bourguignon, Santos-Neto, and de Castro is an alternative to the generalized linear models and useful to model positive asymmetric data. In this paper, we propose two general misspecification tests based on the RESET test for beta prime regression models with varying precision. In the first test, we add the testing variable in the mean submodel, whereas the second test focuses on adding the testing variables in all submodels. We conduct an extensive Monte Carlo simulation study to evaluate the performance of the proposed tests in finite sample size in terms of their sizes and powers, thus obtaining information about the best combination of test statistics and testing variables to perform the proposed tests. We also present and discuss two empirical applications to show the applicability and importance of the proposed tests.

Keywords:

• Misspecification test
• Beta prime regression model
• Incorrect link function
• Monte Carlo simulation
• Omitted variables