2020/09/03 by Thilo Welz, Philipp Doebler, Welz, Thilo +3 · 1 citation
Decision Sciences · Economics, Econometrics and Finance · Medicine · #62F25 (Primary) #62H20 #Diverse Approaches in Healthcare and Education Studies #Economic and Environmental Valuation #FOS: Computer and information sciences #Meta-analysis and systematic reviews #Methodology (stat.ME)
paper · pdf · doi:10.48550/arxiv.2009.01522
openalex publication_date 2020/09/03 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Meta-analyses of correlation coefficients are an important technique to\nintegrate results from many cross-sectional and longitudinal research designs.\nUncertainty in pooled estimates is typically assessed with the help of\nconfidence intervals, which can double as hypothesis tests for two-sided\nhypotheses about the underlying correlation. A standard approach to construct\nconfidence intervals for the main effect is the Hedges-Olkin-Vevea Fisher-z\n(HOVz) approach, which is based on the Fisher-z transformation. Results from\nprevious studies (Field, 2005; Hafdahl and Williams, 2009), however, indicate\nthat in random-effects models the performance of the HOVz confidence interval\ncan be unsatisfactory. To this end, we propose improvements of the HOVz\napproach, which are based on enhanced variance estimators for the main effect\nestimate. In order to study the coverage of the new confidence intervals in\nboth fixed- and random-effects meta-analysis models, we perform an extensive\nsimulation study, comparing them to established approaches. Data were generated\nvia a truncated normal and beta distribution model. The results show that our\nnewly proposed confidence intervals based on a Knapp-Hartung-type variance\nestimator or robust heteroscedasticity consistent sandwich estimators in\ncombination with the integral z-to-r transformation (Hafdahl, 2009) provide\nmore accurate coverage than existing approaches in most scenarios, especially\nin the more appropriate beta distribution simulation model.\n