testing significance of two repeated measures correlations

#1
Hey everybody,

I've run into a small issue that I'm unsure how to solve yet.

I did a repeated measures correlation in r (rmcorr) for a pre and post testing data of the same population. If I want to check if the correlations differ significantly from pre to post, I know I can run a fisher z to r test...but is it suitable for repeated measure correlations or just for Pearson's correlations?

To calculate fisher z to r I need the correlation coefficient and the sample size...but with a repeated measures correlation the number of independent samples is way bigger than just the sample size.

Hope this questions makes any sense. Happy to provide more information if needed.

Thanks,
Dan
 

katxt

Active Member
#3
but is it suitable for repeated measure correlations
I don't know, but I have the same niggles as you do. Perhaps Karabiners reference will tell you if it's OK.
However, you could do a bootstrap test which avoids the question.
You have one pair of rmcorr's and a difference.
Resample the subjects with replacement and find the new rmcorr and difference. Record.
Repeat many times and get a 95% CI for the differences.
Does this 95% CI include zero? If not, then there is a significant difference.