Optimal Pseudo-Gaussian and Rank-Based Random Coefficient Detection in Multiple Regression

Random coefficient regression (RCR) models are the regression versions of random effects models in analysis of variance and panel data analysis. Optimal detection of the presence of random coefficients (equivalently, optimal testing of the hypothesis of constant regression coefficients) has been an open problem for many years. The simple regression case has been solved recently (Fihri et al. (2017)), and the multiple regression case is considered here. This problem poses several theoretical challenges (a)a nonstandard ULAN structure, with log-likelihood gradients vanishing at the null hypothesis; (b) a cone-shaped alternative under which traditional maximin-type optimality concepts are no longer adequate; (c) a matrix of nuisance parameters (the correlation structure of the random coefficients) that are not identified under the null but have a very significant impact on local powers. Inspired by Novikov (2011), we propose a new (local and asymptotic) concept of optimality for this problem, and, for specified error densities, derive the corresponding parametrically optimal procedures.A suitable modification of the Gaussian version of the latter is shown to remain valid under arbitrary densities with finite moments of order four, hence qualifies as a pseudo-Gaussian test. The asymptotic performances of those pseudo-Gaussian tests, however, are rather poor under skewed and heavy-tailed densities. We therefore also construct rank-based tests, possibly based on data-driven scores, the asymptotic relative efficiencies of which are remarkably high with respect to their pseudo-Gaussian counterparts.