[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Fisher's exact test in r x c contingency tables
Date sent: Thu, 2 Jun 2005 07:25:20 +0200
Send reply to: firstname.lastname@example.org
> Dear Listers,
> in my understanding Fisher's exact test (FET) is applicable in
> 2x2 contingency tables only. However, Stata help contends
> that "... Fisher's exact test ... may be applied to r x c tables
> as well as to 2x2 tables".
> I have heard of some algorithm in SAS that seems to apply FET
> to r x c tables, but I still wonder whether this is appropriate. Is
> Stata maybe computing something similar? The only difference in output
> is that in 2x2 tables, 1- and 2-sided p values are computed, whereas
> in r x c tables one (apparently one-sided) p value only is shown.
> What is your opinion? At the moment I rather think about applying
> tests that were designed for ordinal / ordinal relationships in r x c
> contingency tables (say, Goodman and Kruskal's gamma) than embarking
> on a potentially hazardous procedure.
The original refence for the exact test is I believe Example 1 in Fisher, R. A. (1935). The
logic of inductive inference. Journal of the Royal Statistical Society Series A 98, 39–54
(obtainable from the Statistics section of
http://www.library.adelaide.edu.au/digitised/fisher/papers.html which btw has virtually all
of Fisher's papers which make for some interesting reading).
Neither this paper, or Fisher's "Statistical Methods for Research Wrokers" (to the best
of my memory, I don't have a copy to hand at the moment), stipulate that the exact test
is only applicable to 2 x 2 tables, but back then computers weren't so readily available.
I believe the exact tests implemented in Stata are not exactly as the described in the
above references, as the main problem with applying Fisher's exact test to r x c tables
is the exponential increase in computational time required for calculation (for further
details of exactly how the exact p-values are calculated in r x c tables see the
references at the end of the help for -tabulate-).
Personally I don't have a problem with applying the exact test to r x c tables, but I
wouldn't bother trying to apply it to anything with too many dimensions or too large a
sample size, as the computational time is too long (but see comments under exact
option in -tabulate-'s help).
ARC Epidemiology Unit, University of Manchester
"If your result needs a statistician then you should design a better experiment" -
* For searches and help try: