| Sumario: | Meta-analysis is an important tool for identifying best practices. Critical to metaanalysis is the presumption that effect sizes based upon different measurement procedures are directly comparable. Recent theoretical work has challenged this notion, showing that two validity-invariance conditions must hold in order for effect sizes based upon the different measures to be directly comparable. However, no research has been done investigating how much of a practical difference violations of these invariance conditions make in either effect-size variability or in the outcomes of a meta-analysis. This article reports the results of a simulation study of the practical effects that violations of one of these invariance conditions have on the differences in standardized mean difference effect sizes for a given between-population comparison, on the variability in correlation effect sizes for a given relationship in a given population, and on the results of a meta-analysis. Reprinted by permission of the publisher.
|
|---|