--
I think that the manual is misleading here.
The exact-test for RxC tables with R>2 or C>2 is "one-sided", if at
all, in the same way as the Pearson-chi square test is. Both test
the same alternative to the hypothesis of indpendence, but that
alternative is not directional. It is simply "non-independence", or,
as the SAS manual puts it "general association". This is an omnibus
alternative that can include both positive and negative associations
(if row or columns are ordered) and other configurations which have no
"trend" description.
Sidededness makese sense only if there are alternatives in two
directions. This is the case for 2x2 tables: non-independence can be
expressed by saying the OR is not equal to 1. Directional alternatives
are available (OR>1, OR<1), and Fisher's test can be one-sided or
two-sided. Chi square is two-sided.
For R>2 or C>2, I would say that both Fisher and chi square are
"two-sided" or even "multi-sided".
Steve
On Thu, Apr 30, 2009 at 10:34 AM, Richard Williams
<[email protected]> wrote:
> At 08:20 AM 4/30/2009, David Airey wrote:
>>
>> It could be one sided:
>>
>> http://en.wikipedia.org/wiki/Fisher's_exact_test
>>
>> What does the manual say?
>>
>> -Dave
>
> The docs for -tab2- say "For 2 x 2 tables, both one- and two-sided
> probabilities are displayed. For r x c tables, one-sided probabilities are
> displayed."
>
>
> -------------------------------------------
> Richard Williams, Notre Dame Dept of Sociology
> OFFICE: (574)631-6668, (574)631-6463
> HOME: (574)289-5227
> EMAIL: [email protected]
> WWW: http://www.nd.edu/~rwilliam
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
