Abstract
Methods for randomeffects metaanalysis require an estimate of the betweenstudy variance, $\tau^2$. The performance of estimators of $\tau^2$ (measured by bias and coverage) affects their usefulness in assessing heterogeneity of studylevel effects, and also the performance of related estimators of the overall effect. For the effect measure mean difference (MD), we review five point estimators of $\tau^2$ (the popular methods of DerSimonianLaird, restricted maximum likelihood, and Mandel and Paule (MP); the lessfamiliar method of Jackson; and a new method (WT) based on the improved approximation to the distribution of the $Q$ statistic by \cite{kulinskaya2004welch}), five interval estimators for $\tau^2$ (profile likelihood, Qprofile, Biggerstaff and Jackson, Jackson, and the new WT method), six point estimators of the overall effect (the five related to the point estimators of $\tau^2$ and an estimator whose weights use only studylevel sample sizes), and eight interval estimators for the overall effect (five based on the point estimators for $\tau^2$, the HartungKnappSidikJonkman (HKSJ) interval, a modification of HKSJ, and an interval based on the samplesizeweighted estimator). We obtain empirical evidence from extensive simulations and an example.
Original language  English 

Journal  ArXiv eprints 
Publication status  Published  1 Apr 2019 
Keywords
 stat.ME
Profiles

Elena Kulinskaya
 School of Computing Sciences  Emeritus Professor
 Norwich Epidemiology Centre  Member
 Data Science and AI  Member
Person: Honorary, Research Group Member