Small-Sample Adjustments to Multiparameter Hypothesis Tests Based on Robust Variance Estimation in Metaregression

• In Event: 46.081 - Methodological Issues and Meta-Analysis

Abstract

Research syntheses often include studies that report multiple effect size estimates. Meta-analyzing these effect sizes can be difficult because the correlation structure of these effect sizes is often unknown. A recent solution to this problem is robust variance estimation (RVE), which allows dependent effect size estimates to be meta-analyzed without requiring knowledge of the correlation structure. However, corrections to RVE are needed when the number of independent studies is small. Building on previous work for single-parameter t-tests, we investigate small-sample corrections for multi-parameter hypothesis tests (F-tests) using RVE. We develop three tests that have closer to nominal Type I error rates than the conventional asymptotic test. We use simulation methods to compare the performance of the tests.

Authors

• James Eric Pustejovsky, The University of Texas - Austin

• Elizabeth Tipton, Teachers College, Columbia University