Helle Sørensen, KVL
	
Small sample distribution of the LR test in a random effects model

Abstract

We consider a class of random effects model and examine the small
sample distribution of the likelihood ratio test statistic.  The
distribution turns out to depend strongly on the (unknown)
heterogeneity parameter of the model.  We consider a related
overdispersion model in order to understand this dependence.  The
random effects models are often used in meta-analysis where results
from (a few) related but independent studies are combined in order to
achieve an estimate across studies. We will use such a meta-analysis,
as well as an example with mixtures of pesticides, as illustrations.