zbMATH — the first resource for mathematics

Consistent specification testing via nonparametric series regression. (English) Zbl 0941.62125
Summary: This paper proposes two consistent one-sided specification tests for parametric regression models, one based on the sample covariance between the residuals from the parametric model and the discrepancy between the parametric and nonparametric fitted values, the other based on the difference in sums of squared residuals between the parametric and nonparametric models. We estimate the nonparametric model by series regression. The new test statistics converge in distribution to a unit normal under correct specification and grow to infinity faster than the parametric rate \((n^{-1/2})\) under misspecification, while avoiding weighting, sample splitting, and non-nested testing procedures used elsewhere in the literature. Asymptotically, our tests can be viewed as a test of the joint hypothesis that the true parameters of a series regression model are zero, where the dependent variable is the residual from the parametric model, and the series terms are functions of the explanatory variables, chosen so as to support nonparametric estimation of a conditional expectation.
We specifically consider Fourier series and regression splines, and present a Monte Carlo study of the finite-sample performance of the new tests in comparison to consistent tests of H.J. Bierens [ibid. 58, No. 6, 1443-1458 (1990; Zbl 0737.62058)], R.L. Eubank and C.H. Spiegelman [J. Am. Stat. Assoc. 85, No. 410, 387-392 (1990; Zbl 0702.62037)], J.M. Wooldridge [Int. Econ. Rev. 33, No. 4, 935-955 (1992; Zbl 0781.62100)], and some others. The results show the new tests have good power, performing quite well in some situations. We suggest a joint Bonferroni procedure that combines the new test with those of Bierens and Wooldridge to capture the best features of the three approaches.

62P20 Applications of statistics to economics
62G08 Nonparametric regression and quantile regression
Full Text: DOI