[R-meta] pooled vs separate QB statistics
Dear Michael, Thank you. So basically when different tau2 are estimated (either using a multivariate model or scale location model) the Q test for the moderators is the non-pooled test? When a single tau2 is estimated the Q test is the pooled version? Filippo
On Tue, 28 May 2024 at 12:31, Michael Dewey <lists at dewey.myzen.co.uk> wrote:
Dear Filippo Does https://www.metafor-project.org/doku.php/tips:different_tau2_across_subgroups help? Michael On 28/05/2024 11:09, Filippo Gambarota via R-sig-meta-analysis wrote:
Hi, I was reading the two papers (at the end) about categorical moderators and the omnibus test and despite the conclusion is that the Q (or F) test based on the pooled tau2 estimation works better in most cases, when k in each moderator level is large and tau2 is expected to differ across levels, it is better to use the non-pooled method. Reading the metafor documentation it is not clear to me how to implement the non-pooled method. Maybe I am missing something but I can change the type of test (z vs t vs knha that takes into account tau2 uncertainty) but I see no option to choose the pooled vs non pooled tau2 method. Thanks Filippo Rubio-Aparicio, M., L?pez-L?pez, J. A., Viechtbauer, W., Mar?n-Mart?nez, F., Botella, J., & S?nchez-Meca, J. (2020). Testing Categorical Moderators in Mixed-Effects Meta-analysis in the Presence of Heteroscedasticity. Journal of Experimental Education, 88(2), 288?310. https://doi.org/10.1080/00220973.2018.1561404 Rubio-Aparicio, M., S?nchez-Meca, J., L?pez-L?pez, J. A., Botella, J., & Mar?n-Mart?nez, F. (2017). Analysis of categorical moderators in mixed-effects meta-analysis: Consequences of using pooled versus separate estimates of the residual between-studies variances. The British Journal of Mathematical and Statistical Psychology, 70(3), 439?456. https://doi.org/10.1111/bmsp.12092
-- Michael
Filippo Gambarota, PhD Postdoctoral Researcher - University of Padova Department of Developmental and Social Psychology Website: filippogambarota.xyz Research Groups: Colab Psicostat