1

I would like to test the difference between 2 correlations on statistical significance by using confidence intervals.

Regarding means: By now I found out, that when comparing 2 means overlapping/non-overlapping 84% confidence intervals are close to a significance test at 5%, while computing 95% confidence intervals will be to conservative:

on stack exchange: Family-wise confidence intervals or ANCOVA in R suggests different intercepts, but the 95% CIs overlap... how is this possible? And here is a paper i found on google scholar regarding the comparision of means using an 84% confidence interval: https://academic.oup.com/jinsectscience/article/2577125

However, i couldn´t find anything regarding correlations.

So my question: Can i use a 84% confidence interval for testing correlations, just like it seems to be done with means?

bucky
  • 205
  • 2
  • 11
  • 1
    Perhaps this post https://stats.stackexchange.com/a/73628/128491 might be of help if I am understanding your question correctly. – Just_to_Answer Aug 05 '17 at 16:53
  • 1
    That approach does not generalize from means to correlations. Why not compute the 95% confidence interval on the difference? Or, just test the significance of the difference [directly](http://vassarstats.net/rdiff.html). – David Lane Aug 05 '17 at 18:42
  • thank you both for your answers. Mr. Lane answered my question by telling, that the 84% CI rule is not applicable for correlations. I will therefore simply calculate the 95% CI. – bucky Aug 06 '17 at 00:41

0 Answers0