[Rd] cor.test(x,x) sometimes produce NaN (PR#1099)

maechler@stat.math.ethz.ch maechler@stat.math.ethz.ch
Mon, 24 Sep 2001 14:00:06 +0200 (MET DST)

This *is* kind of a bug, hence diverted to R-bugs

>>>>> "Peter" == Peter Malewski <peter.malewski@gmx.de> writes:

    Peter> In the code

    >> for(i in 1:100){zz <- rnorm(10);print(cor.test(zz,zz))}

    Peter> sometimes a sqrt(0) occures:
    Peter> 24: NaNs produced in: sqrt(1 - r^2)

I found that the reason is that in (at least one of) those cases,
  r == 1+eps
where eps = .Machine$double.eps
which I think is the real bug, here.

Here is a reproducible example:

 > set.seed(125);z0 <- rnorm(40);zz <- round(rnorm(10),5); r <- cor(zz,zz)
 > r-1
 [1] 2.220446e-16

    Peter> in all the other cases a result like
(for me, not in all, but quite a few cases)

    Peter>   t = Inf, df = 8, p-value = < 2.2e-16

which I think is entirely correct.

    Peter> occurs. Does cor.test need some more "inf" - handling?  I found
    Peter> that calculating a cor-matrix with additional p-values.  
  (last sentence is a bit unclear to MM).

Martin Maechler <maechler@stat.math.ethz.ch>	http://stat.ethz.ch/~maechler/
Seminar fuer Statistik, ETH-Zentrum  LEO D10	Leonhardstr. 27
ETH (Federal Inst. Technology)	8092 Zurich	SWITZERLAND
phone: x-41-1-632-3408		fax: ...-1228			<><

r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch