[R] Correlation with small N

[Jonathan Baron]

On 05/17/02 11:03, Christian Schulz wrote:
>Hello r-proffessionals,
>this is not a special R-Question, but have anybody
>a good reason for nonstatistican's why the interpretation of correlations  
>with cases under 20-30 are dangerous !?

I am not an R-professional, nor a statistician, but this is
off-topic anyway, so ...

I don't think it is especially true that interpretation of
correlations with small N is dangerous.  At least it is no more
dangerous than interpretation of any statistic with small N.

There is an issue here, I think, when small N is combined with
the practice of reporting p-levels by setting an "alpha" cutoff
for significance (like p<.05) and then reporting only whether p
is below the cutoff.  The exact p is hidden by such a practice.
If all you know is that p<.05 and that N=20, then the expectation
of the exact p is higher than if p<.05 and N=200.  So sample size
does matter even when alpha (the cutoff) is held constant.

I do not think that this problem arises when exact p levels are
reported.  In psychology, exact p levels are more common now, I
think, and they are actually recommended by the American
Psychological Association: 
http://www.apa.org/journals/amp/amp548594.html

Perhaps there is something else here that I don't know.

-- 
Jonathan Baron, Professor of Psychology, University of Pennsylvania
Home page: http://www.sas.upenn.edu/~baron
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help 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-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._