[R-meta] Significance of the individual effects
Viechtbauer, Wolfgang (SP)
wolfg@ng@viechtb@uer @ending from m@@@trichtuniver@ity@nl
Tue Sep 25 17:32:01 CEST 2018
If you just want to test, then:
2*pnorm(abs(yi)/sqrt(vi), lower.tail=FALSE) <= .05
will tell you which estimates are significantly different from 0 at alpha=.05 (two-sided), assuming (approximate) normality of the sampling distribution.
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On Behalf Of Tommy van Steen
Sent: Tuesday, 25 September, 2018 1:25
To: Cesar Terrer Moreno
Cc: r-sig-meta-analysis using r-project.org
Subject: Re: [R-meta] Significance of the individual effects
I’m not sure why you do the ‘exp’ bit, as removing that bit gives the right lower limits compared to a simple forest plot produced using the metafor-package.
Your ifelse function will only work if all your effect sizes are positive, so it should work in your example, but you cannot use it to decide whether negative effect sizes are significant or not (You’d need the opposite, whether the yi + upper limit >= 0). If you have many studies, a forest plot where you rank the studies by 95%CI lower limit might be an easier way to inspect significance of all individual studies.
> On 24 Sep 2018, at 17:46, Cesar Terrer Moreno <cesar.terrer using me.com> wrote:
More information about the R-sig-meta-analysis