Your input and results


This gives you all four recognised ways of getting a confidence interval around an observed Spearman correlation. The four options come from two binary choices. The first is whether to look up the coverage quantile from the t distribution (with n - 2 degrees of freedom) which is the default, or from the Gaussian distribution.

The other option is whether to use the equation of Bonett & White (2000) for the SE of the correlation or to use the equation of Fieller, Hartley & Pearson (1957)

I have set the defaults here to use the Bonett & White method with the t distribution, matching the defaults in the getCISpearman() function in the CECPfuns package

Both methods are approximations and all four approaches are known give poor coverage for n < 25 or for very strong correlations (i.e. absolute correlation > .9) Having said that, the methods are moderately robust for n > 25 and abs(rs) < .9 and if all you have is the observed correlation and the n it's as good as you can get. If you have the raw data I recommend you use a bootstrap CI of the correlation. I'll put up an app to do that when I can.

App created by Chris Evans licenced under a Creative Commons, Attribution Licence-ShareAlike Please respect that and put an acknowledgement and link back to here if re-using anything from here.

Background and related resources

This shiny app is one of a growing number in, my shiny server
They complement:

There is a form if you want to contact me: so do please use that if you think there is anything wrong here or anything that could be improved.

There is also now an Email announcement list, never updating more than monthly, where I will put up developments of new apps here, a summary of updates to the glossary and new posts in the Rblog. You can sign up for that here.