Tue Feb 12 06:43:05 CST 2013
On 12 February 2013 13:19, Jerome Kieffer <Jerome.Kieffer@esrf.fr> wrote:
> I am not very confident in my knowledge in statistics but here R = -1 and it does not look like an error.
> I implemented a method to perform thousands of linear regression and few of them returned -1.00001 (or so) which later gave NaN as stderr.
You are absolutely right, my mistake. I got the formula confused in my
head, I should have looked it up. Indeed, the covariance can be
negative, giving you the r<0.
More information about the SciPy-User