[R] corr.test -- use a different null hypothesis

Dieter Menne dieter.menne at menne-biomed.de
Fri Sep 16 13:00:41 CEST 2005


Dr. Hans Georg Krauthaeuser <Hans-Georg.Krauthaeuser <at> E-Technik.Uni-
Magdeburg.DE> writes:

> 
> I want to use corr.test to test for the correlation of two data sets
> (actually I have a lot of data set and perform pairwise testing). But I
> wanted to find sets where the correlation is less than a certain limit,
> i.e. |r_exp| < r_0. As far as I can see corr.test is designed to work
> only for the null hypothesis r_exp = 0. I would need H0 as |r_exp|>= r_0.
> 

cor.test (sic!) for method="pearson" returns a confidence interval for r. You 
could use the upper limit of it for your test

Dieter




More information about the R-help mailing list