Timeline for How robust is Pearson's correlation coefficient to violations of normality?
Current License: CC BY-SA 3.0
8 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 26, 2014 at 12:18 | answer | added | Karl Ove Hufthammer | timeline score: 24 | |
| Sep 30, 2012 at 4:33 | comment | added | Archaeopteryx | I'm just using it for simple bivariate correlations. I'm not sure why it is claimed that normality is required. The stats texts I've read always list normality as an assumption of the Pearson's correlation and advise to use Spearman's for conditions in which non-normality holds. | |
| Sep 30, 2012 at 4:11 | comment | added | Douglas Zare | Spearman's rank correlation coefficient is Pearson's correlation coefficient applied to the non-normal rankings. I still don't know in what sense you believe Pearson's requires normality. Perhaps you can say a few extra things in case you are using it on a multivariate normal distribution. | |
| Sep 30, 2012 at 2:31 | comment | added | Archaeopteryx | That Pearson's correlation assumes normality is what many stats texts claim. I've heard elsewhere that normality is a needless assumption for Pearson's r. When I run the analyses, both Pearson's and Spearman's produce relatively similar results. | |
| Sep 29, 2012 at 21:04 | comment | added | Douglas Zare | Wait, Pearson's correlation coefficient assumes normality? I don't think it does, and I have been using it on non-normal data. It's just not robust to some things which happen more often in some non-normal situations, but there are plenty of non-normal situations where I see no problem with using Pearson's correlation coefficient. | |
| Sep 29, 2012 at 21:01 | history | tweeted | twitter.com/#!/StackStats/status/252151037052469248 | ||
| Sep 29, 2012 at 14:38 | answer | added | Peter Flom | timeline score: 5 | |
| Sep 29, 2012 at 14:32 | history | asked | Archaeopteryx | CC BY-SA 3.0 |