infotest: Information Matrix Test for Linear Regression Models in R

Abstract This article introduces the infotest package for R, which implements the Information Matrix (IM) test for linear regression models. The IM test, originally proposed by White (1982) and later decomposed by Cameron and Trivedi (1990), provides a comprehensive diagnostic tool for assessing model misspecification. Building on the results of Chesher (1983) and Lancaster (1984), the test examines whether the information matrix equality holds–a fundamental property of correctly specified maximum likelihood models. The decomposition into heteroskedasticity, skewness, and kurtosis components allows researchers to identify specific sources of misspecification. Unlike unconditional normality tests, the IM test provides conditional moment testing that accounts for covariate patterns. The packagealsoincludes White’s classic heteroskedasticityt est (White1980) as a special case. We demonstrate the implementation through theoretical background, computational details, and practical examples.