Capturing study influence: the concept of 'gravity' in meta-analysis

Abstract

While much theory abounds over properties of meta-analyses, there appears to be very little work to date on examining jackknifed and exact distributions of the statistics generated by the method. This paper takes an initial step towards filling that gap by describing a SAS (SAS Institute, 2001) macro written by the author based on another macro due to others, that performs jackknifed estimates of average effect size. It also suggests that ‘gravity’ is a property of studies included in a meta-analysis. A computer simulation supports the utility of the approach, and proposals for future development of exact and approximate methods in ‘combinatorial meta analysis’ are set forward