cor.test observations limit - R-help

Tue, Apr 1, 2003 10:57 AM #

Hi,

Is there a limit on the number of observations for using cor.test. For
example,

Error in if (q > (n^3 - n)/6) pspearman(q - 1, n, lower.tail = FALSE) else
pspearman(q,  : 
        missing value where logical needed
In addition: Warning message: 
NAs introduced by coercion 

I mainly want to calculate the spearman correlation rho for a data with
missing values. But the rcorr function in Hmisc library (from
Hmisc_1.4.2.zip) gave me an error message

Error in rcorr(tmp[5600:5900, 1:2], type = "spearman") : 
        NA/NaN/Inf in foreign function call (arg 1)

Any help is appreciated,

Ming-Chung Li

Brian Ripley

Tue, Apr 1, 2003 11:45 AM #

Yes.  About 1290 for the Spearman test.

You need (n^3 - n) representable as an integer.

On Tue, 1 Apr 2003, Ming-Chung Li wrote:

That's not needed, and has not been for about 3 years.

Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Ming-Chung Li

Wed, Apr 2, 2003 10:00 AM #

This is a note to my second question. I just discovered my data containing
an 'Inf' value and this causes the error I saw when I used rcorr() function
from the Hmisc package. If I delete this observation, everything is fine.

Best,
Ming-Chung