A score test for variance components in a semiparametric mixed-effects model under non-normality

Abstract

In this paper, we propose a score test for variance components in a semiparametric mixed-effects model when the random-effects and measurement errors are not normally distributed. The asymptotic null distribution of the test statistic is shown to be a simple chi-squared distribution with the degrees of freedom being the number of linearly-independent variance components. The simulation results show that the proposed score test is robust against the nonnormality of the random-effects and the measurement errors and performs well in terms of both size and power. The score test is illustrated via an application to a real longitudinal data set collected in a clinical trial study.