0

I'm trying to create a forest plot in metafor for pooled reliability estimates (Cronbach's Alpha and ICC) for a systematic review we've done. I want to do the same as the authors have in the image I've pasted (which comes from here: https://doi.org/10.1177/1071100718816953). i.e.: we'd like to recreate these sorts of analyses using our data. I can't find a tutorial or example code for this despite exhausting the search engine here and google, nor can I figure out from the help which doesn't seem to touch on this (https://www.rdocumentation.org/packages/metafor/versions/1.9-9/topics/escalc or https://cran.r-project.org/web/packages/metafor/metafor.pdf). I've emailed the corresponding author who suggested we read: https://www.um.es/metaanalysis/pdf/5049.pdf, but this hasn't helped (nothing about R or the steps involved in creating the analyses/plots in there).

Can anyone point me in the direction of some help for the steps required to do this?

Here's the image:

Figure displaying Forest plot of pooled reliability measures

Edit: clarified we were trying to do our own analysis, not recreate someone else's, & added the corresponding author's reply

Rod Whiteley
  • 13
  • 4
  • Doesn't appear like the supplementary materials in the article provide data, do you have the data? – Chris Jul 20 '20 at 14:25
  • Sorry for the misunderstanding - we are looking to do the same sort of analyses on our own data, but can't figure out how to do this using metafor. If possible does anyone know of a tutorial or example code which goes through the process of creating the Forest plots of reliability as in the example above (so we can do the same with our data)? – Rod Whiteley Jul 21 '20 at 06:56

1 Answers1

1

Here is an example for Cronbach's alpha, using the transformation by Bonett (2002) before analyzing the alpha values (not sure what the authors used).

library(metafor)
dat <- dat.bonett2010
res <- rma(measure="ABT", ai=ai, mi=mi, ni=ni, data=dat)
forest(res, slab=dat$source, header=TRUE, atransf=transf.iabt, refline=coef(res))

forest plot

Adjust as needed. See help(forest.rma) (or here) for details on what can be adjusted.

Wolfgang
  • 2,810
  • 2
  • 15
  • 29
  • Thanks Wolfgang - final question, what measure should be used when pooling an ICC? From the escalc help files it's not clear: Partial and Semi-Partial Correlations ... The options for the measure argument are then: "PCOR" for the partial correlation coefficient. "ZPCOR" for Fisher's r-to-z transformed partial correlation coefficient. "SPCOR" for the semi-partial correlation coefficient. – Rod Whiteley Aug 19 '20 at 08:17
  • An ICC is not a partial or semi-partial correlation. There is no functionality in `escalc()` for dealing with ICCs. – Wolfgang Aug 19 '20 at 15:38
  • Thanks again - it looks like there's no clear method to pool ICC's according to: https://www.sciencedirect.com/science/article/pii/S1053811919307487 & they recommend against Fisher Z transforming as the results of the meta analysis can't be interpreted. I can't figure out what the authors in https://doi.org/10.1177/1071100718816953 did to get their results that were the original motivation here (they haven't been helpful in emails). If I've missed a method, please let me know. If I figure this out, I'll post a reply here. – Rod Whiteley Aug 20 '20 at 07:18