Meta‐Analysis Model
Eugene Demidenko · Wiley series in probability and statistics · 1987
The meta-analysis model is probably the simplest random effects model, yet it has important applications for pooling studies, especially in medicine and epidemiology. This chapter considers a simple meta-analysis model where likelihood and quadratic unbiased estimation methods are described at length. It describes the meta-analysis model with study-specific covariates to adjust the common treatment effect. The chapter presents a general multivariate meta-analysis model. Some efficiency may be gained if the treatment effect is observed along with other, correlated estimates. The results of multivariate meta-analysis model may be applied to the multiresponse weighted regression model. The robust/median meta-analysis model assumes that the random effect has Laplace distribution with heavy tails. This model may be useful when studies are heterogeneous or in the presence of outlier (s). Controlled Vocabulary Terms covariance matrix; covariate; Laplace distribution; multivariate statistics