An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20120802/10315fa5/attachment.pl>
Metafor package: Including multiple (categorical) predictors
3 messages · Bexkens, Anika, Jeremy Miles, Viechtbauer Wolfgang (STAT)
The test of moderator coefficients (QM) is chi-square distributed.You can use the change in this value when you add a predictor to the model as a chi-square test, with df equal to the change in df. Jeremy
On 2 August 2012 05:54, Bexkens, Anika <A.Bexkens at uva.nl> wrote:
Dear Metafor users,
I'd like to test a model with 2 continuous and 2 categorical moderators in a meta regression. One categorical parameter has 2 levels and the other has 4 levels. If I understand correctly, when I include all moderators in the model, Metafor returns main effects of the continuous parameters and contrasts of each level of categorical moderators with the intercept (which includes the reference level of the categorical parameters).
This makes it possible to see whether different levels of the categorical moderator are differentially related to effect size. I include multiple moderators and would like to report for each variable whether it is significantly moderating effect size. Is it possible to obtain an overall main effect of each categorical variable, instead of the contrast effects? Or can I only obtain this by including one categorical moderator at a time and reporting the omnibus moderator test?
Many thanks,
Anika
[[alternative HTML version deleted]]
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Just to follow up on that: You can use the 'btt' argument in the rma() function to specify which coefficients to include in the QM test. For example: data(dat.bcg) dat <- escalc(measure="RR", ai=tpos, bi=tneg, ci=cpos, di=cneg, data=dat.bcg, append=TRUE) rma(yi, vi, mods = ~ factor(alloc) + year + ablat, data=dat, btt=c(2,3)) will give you a test of the alloc factor. Note that it does not matter which level of the factor is the reference level: rma(yi, vi, mods = ~ relevel(factor(alloc), ref="random") + year + ablat, data=dat, btt=c(2,3)) This will give you a Wald-type test. Alternatively, you can use a likelihood ratio test (for this, you have to use method="ML"): res1 <- rma(yi, vi, mods = ~ factor(alloc) + year + ablat, data=dat, method="ML") res0 <- rma(yi, vi, mods = ~ + year + ablat, data=dat, method="ML") anova(res1, res0) Best, Wolfgang -- Wolfgang Viechtbauer, Ph.D., Statistician Department of Psychiatry and Psychology School for Mental Health and Neuroscience Faculty of Health, Medicine, and Life Sciences Maastricht University, P.O. Box 616 (VIJV1) 6200 MD Maastricht, The Netherlands +31 (43) 388-4170 | http://www.wvbauer.com
-----Original Message----- From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf Of Jeremy Miles Sent: Friday, August 03, 2012 04:01 To: Bexkens, Anika Cc: r-help at r-project.org Subject: Re: [R] Metafor package: Including multiple (categorical) predictors The test of moderator coefficients (QM) is chi-square distributed.You can use the change in this value when you add a predictor to the model as a chi-square test, with df equal to the change in df. Jeremy On 2 August 2012 05:54, Bexkens, Anika <A.Bexkens at uva.nl> wrote:
Dear Metafor users, I'd like to test a model with 2 continuous and 2 categorical moderators
in a meta regression. One categorical parameter has 2 levels and the other has 4 levels. If I understand correctly, when I include all moderators in the model, Metafor returns main effects of the continuous parameters and contrasts of each level of categorical moderators with the intercept (which includes the reference level of the categorical parameters).
This makes it possible to see whether different levels of the
categorical moderator are differentially related to effect size. I include multiple moderators and would like to report for each variable whether it is significantly moderating effect size. Is it possible to obtain an overall main effect of each categorical variable, instead of the contrast effects? Or can I only obtain this by including one categorical moderator at a time and reporting the omnibus moderator test?
Many thanks, Anika