Sample Size Calculation for Diagnostic Tests in Generalized Linear Mixed Models

Abstract

Intra-cluster correlations have to be taken into account for calculating the stochastic behavior of estimators in diagnostic studies with repeated measurements in individuals. One approach of inducing the intra-cluster correlation is provided by generalized linear mixed models. The inverse of the Fisher information matrix as the asymptotic covariance of the maximum likelihood estimator is in these models of interest for determining the required sample size in statistical tests.We illustrate the dependence of the sample size on different approximations of the Fisher information matrix for the prove of non-inferiority in diagnostic studies.