[S] strange p-value for Fisher exact's test

Marcia Ciol (marcia@biostat.washington.edu)
Mon, 23 Mar 1998 11:53:08 -0800

I came across a p-value greater than one in a Fisher exact's test. The
output was the following

> fisher.test(rbind(c(1, 1), c(71, 77)))

Fisher's exact test

data: rbind(c(1, 1), c(71, 77))
p-value = 1.0001
alternative hypothesis: two.sided

I know I could just look at the table values to see that the p-value is
1.0, but it worries me that the internal programing of S-Plus would
allow the result above to occur. SPSS, for example, will give the
correct value (at least up to 17 decimal digits)...

Anybody would like to comment on the subject?


Marcia A. Ciol                  mailto:marcia@biostat.washington.edu 
Research Scientist                        
                                          University of Washington
Center for Cost and Outcomes Research     Department of Health Services
146 N Canal Street, Suite 300             UW Campus Mail: Box 358853
Seattle   WA  98103-8652                  Fax    (206) 685-3467
                                          Office (206) 543-4887
This message was distributed by s-news@wubios.wustl.edu.  To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message:  unsubscribe s-news