On the Power of Invariant Tests for Hypotheses on a Covariance Matrix

TL;DR

This paper develops a comprehensive theory for invariant tests of covariance matrices in linear regression, explaining their power behavior under strong correlations and identifying conditions where null and alternative hypotheses are indistinguishable.

Contribution

It constructs a general framework for invariant tests on covariance matrices, addressing previous flaws and extending the theory to spatial and time series autocorrelation testing.

Findings

Power of invariant tests can converge to zero under strong correlation.

Characterizes when null and alternative hypotheses are indistinguishable.

Provides a unified theory covering spatial and time series models.

Abstract

The behavior of the power function of autocorrelation tests such as the Durbin-Watson test in time series regressions or the Cliff-Ord test in spatial regression models has been intensively studied in the literature. When the correlation becomes strong, Kr\"amer (1985) (for the Durbin-Watson test) and Kr\"amer (2005) (for the Cliff-Ord test) have shown that the power can be very low, in fact can converge to zero, under certain circumstances. Motivated by these results, Martellosio (2010) set out to build a general theory that would explain these findings. Unfortunately, Martellosio (2010) does not achieve this goal, as a substantial portion of his results and proofs suffer from serious flaws. The present paper now builds a theory as envisioned in Martellosio (2010) in a fairly general framework, covering general invariant tests of a hypothesis on the disturbance covariance matrix in a linear regression model. The general results are then specialized to testing for spatial correlation and to autocorrelation testing in time series regression models. We also characterize the situation where the null and the alternative hypothesis are indistinguishable by invariant tests.