Meta-analysis and evidence

https://doi.org/10.1016/0197-2456(89)90030-5Get rights and content
Under a Creative Commons license
open archive

Abstract

Meta-analysis is the science of combining evidence from different studies, but traditional statistical techniques contain neither a formal definition nor a measure of evidence. It is argued in this paper that the log-likelihood ratio, as a measure of the “weight of evidence” can be a very useful tool in the meta-analysis. The mathematics and the philosophy behind the use of this index are introduced. The construction and interpretation of “support curves” in fixed and random-effects models are presented. The application of evidential techniques is illustrated on six trials of aspirin therapy previously presented by Canner. The possible dangers of focusing on statistical error rates instead of evidence are discussed.

Keywords

Inference
evidence
p values
likelihood
ratio
support curves
pooling
meta-analysis

Cited by (0)

Parts of this paper were presented at the meeting of The Society for Clinical Trials, Atlanta, GA, in May 1987.