Our study revisits and challenges two core conventional meta-regression models: the prevalent use of ?mixed-effects? or random-effects meta-regression analysis (RE-MRA) and the correction of standard errors that defines fixed-effects meta-regression analysis (FE-MRA). We show how and explain why the traditional, unrestricted weighted least squares estimator (WLS-MRA) is superior to conventional random-effects (or mixed-effects) meta-regression when there is publication (or small-sample) bias and as good as FE-MRA in all cases and better in most practical applications. Simulations and statistical theory show that WLS-MRA provides satisfactory estimates of meta-regression coefficients with confidence intervals that are comparable to mixed-or random-effects when there is no publication bias. When there is publication selection bias, WLS-MRA dominates mixed- and random-effects, especially when there is large additive heterogeneity as assumed by the random-effects meta-regression model.
History
Series
School Working Paper - Economics Series ; SWP 2013/2
Pagination
1 - 26
Publisher
Deakin University, School of Accounting, Economics and Finance
Place of publication
Geelong, Vic.
Language
eng
Notes
School working paper (Deakin University. School of Accounting, Economics and Finance) ; 2013/2
Our study revisits and challenges two core conventional meta-regression models: the prevalent use of ?mixed-effects? or random-effects meta-regression analysis (RE-MRA) and the correction of standard errors that defines fixed-effects meta-regression analysis (FE-MRA). We show how and explain why the traditional, unrestricted weighted least squares estimator (WLS-MRA) is superior to conventional random-effects (or mixed-effects) meta-regression when there is publication (or small-sample) bias and as good as FE-MRA in all cases and better in most practical applications. Simulations and statistical theory show that WLS-MRA provides satisfactory estimates of meta-regression coefficients with confidence intervals that are comparable to mixed-or random-effects when there is no publication bias. When there is publication selection bias, WLS-MRA dominates mixed- and random-effects, especially when there is large additive heterogeneity as assumed by the random-effects meta-regression model.