Converting spearmans r to fishers z for meta analysis ; Correcting for range in meta analysis

[laboh]

Hi all

I am conducting a meta analysis of pearsons r correlations but I have a study with a spearmans rank correlation. I was converting the pearsons correlations into fishers z scores to conduct the meta analysis , then to be converted back into correlations, but I am wondering if I am able to do this with a spearmans rank correlation? Can I use the same calculator as for a pearsons r > fishers z score?

Also in order to correct for range restriction within all the data from various articles , is there some form of data I should be seeking from those articles or can I calculate this in some other way?

Thanks

 

[hlsmith]

Good question. I have never thought about this one. I wonder if all of the other studies used Pearson's, that the normality isn't too off in the underlying data. Given this I also wonder if you can find any papers addressing how comparable Pearson and Spearman corrs are to each other in near normality settings. If they are comparable I may just think about using the same conversion. When doing your sensitivity analysis on the meta-analysis (leave-one-out) you will be able to see how big of an effect it may have on the overall analyses. Also, if the study has little weight over all in the MA, given say a small sample, having just one spearman rank in the mix may not be a big deal. It seems like they have a fisher conversion for everything so keep digging around and report back if you find anything substantive. If you find a fisher conversion, you could convert it both ways and see if it makes any difference.

Also, if you were obsessed with the question you could also run a simulation study to see if Pearson and Spearman are comparable enough in the setting and what effect it may have on estimates.

Good luck.