[R] metafor - multivariate analysis

[Viechtbauer Wolfgang (STAT)]

Dear Owen,

What is your definition of "multivariate analysis"? Do you mean: A meta-regression model with more than one predictor/moderator? In that case, yes, metafor handles that. Usually, this is referred to as "multiple regression" (as opposed to "simple regression" with a single predictor) -- and in the case of a meta-analysis, I guess one could call it "multiple meta-regression". 

If you are referring to a model that handles statistical dependencies in the observed outcomes (and hence requires multivariate methods), then you will have to use some other package (e.g., mvmeta).

See also:

http://stats.stackexchange.com/questions/2358/explain-the-difference-between-multiple-regression-and-multivariate-regression

Best,
Wolfgang

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
> On Behalf Of Owen, Branwen
> Sent: Friday, March 15, 2013 11:23
> To: r-help at r-project.org
> Subject: [R] metafor - multivariate analysis
> 
> 
> 
> 
> Dear Metafor users, I'm conducting a metaanalysis of prevalence of a
> particular behaviour based on someone elses' code. I've been labouring
> under the impression that this:
> 
> summary(rma.1<-
> rma(yi,vi,mods=cbind(approxmeanage,interviewmethodcode),data=mal,method="D
> L",knha=F,weighted=F,intercept=T))
> 
> is doing the multivariate analysis that i want, but have read that
> multivariate analysis can't be done in metafor.
> 
> this is the output:
> 
> Mixed-Effects Model (k = 22; tau^2 estimator: DL)
> 
> logLik Deviance AIC BIC
> 18.7726 -37.5452 -27.5452 -22.0899
> 
> tau^2 (estimate of residual amount of heterogeneity): 0.0106
> tau (sqrt of the estimate of residual heterogeneity): 0.1031
> 
> Test for Residual Heterogeneity:
> QE(df = 18) = 1273.9411, p-val < .0001
> 
> Test of Moderators (coefficient(s) 2,3,4):
> QM(df = 3) = 11.0096, p-val = 0.0117
> 
> Model Results:
> 
> estimate se zval pval ci.lb ci.ub
> intrcpt 0.4014 0.1705 2.3537 0.0186 0.0671 0.7356 *
> continent -0.0206 0.0184 -1.1200 0.2627 -0.0568 0.0155
> approxmeanage 0.0076 0.0091 0.8354 0.4035 -0.0102 0.0254
> interviewmethodcode -0.0892 0.0273 -3.2702 0.0011 -0.1426 -0.0357 **
> 
> ---
> Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
> 
> my questions are
> 1. what is this line of code?
> 2. if it isn't multivariate analysis, will i have to use the mvmeta
> instead.
> 
> thanks very much for any help
> Branwen
> .<http://r.789695.n4.nabble.com/metafor-multivariate-analysis-
> td4661233.html>
> 
> 
> 	[[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.