Fisher's Exact Test for 3x2 Table ?
Tan Teng Hong
I have only access to SPSS 6.1 in my institution, and SPSS 6.1 does
not give Fisher's exact test for tables larger than 2x2.
I have the following table, and need to determine whether Group A and
Group B are significantly different in terms of X,Y,Z. Expected
values in some cells are < 5.
! A B
--+--------------
X ! 7 9
Y ! 11 9
Z ! 2 2
I will be very greatful if someone can help me. Thanks in advance.
Tan Teng Hong
tan...@merlion.singnet.com.sg
David Nichols
In article <33b4e2bc...@news.singnet.com.sg>,
SPSS with just the Base module doesn't produce a Fisher exact result
for tables larger than 2x2 for any release. The Exact Tests module,
which is available for Releases 6.1.2 and above, offers the Fisher
for RxC tables.
--
-----------------------------------------------------------------------------
David Nichols Senior Support Statistician SPSS, Inc.
Phone: (312) 329-3684 Internet: nic...@spss.com Fax: (312) 329-3668
-----------------------------------------------------------------------------
Bruce Weaver
On Sat, 28 Jun 1997, Tan Teng Hong wrote:
> I have only access to SPSS 6.1 in my institution, and SPSS 6.1 does
> not give Fisher's exact test for tables larger than 2x2.
>
> I have the following table, and need to determine whether Group A and
> Group B are significantly different in terms of X,Y,Z. Expected
> values in some cells are < 5.
>
>
> ! A B
> --+--------------
> X ! 7 9
> Y ! 11 9
> Z ! 2 2
>
> I will be very greatful if someone can help me. Thanks in advance.
>
> Tan Teng Hong
> tan...@merlion.singnet.com.sg
I don't think you really need a statistical test to see that
groups A and B do not differ. But if I needed a test, I would just
perform a chi-square test of independence, but use the likelihood
chi-square rather than Pearson's chi-square. According to Hays (1963),
there is reason to believe that the likelihood chi-square is less affected
by small sample sizes (and small expected frequencies) than Pearson's
chi-square, particularly when df > 1. For these data, the likelihood
chi-square (with 2 df) = 0.451, which is not significant.
Bruce Weaver
UWB, Psychology
pss...@bangor.ac.uk
mrg...@nyu.edu
> I have only access to SPSS 6.1 in my institution, and SPSS 6.1 does
> not give Fisher's exact test for tables larger than 2x2.
>
> I have the following table, and need to determine whether Group A and
> Group B are significantly different in terms of X,Y,Z. Expected
> values in some cells are < 5.
>
>
> ! A B
> --+--------------
> X ! 7 9
> Y ! 11 9
> Z ! 2 2
>
> I will be very greatful if someone can help me. Thanks in advance.
>
>
>
> Tan Teng Hong
> tan...@merlion.singnet.com.sg
Rich Ulrich
A note from 1997?
1) I don't remember whether SPSS 6.1 did provide other
Exact Tests, but Exact Tests have been available for a
long time now, in SPSS. It was first presented as part of
a special procedure for EXACT.
2) By the way, the FET, by that precise name, does not
exist/ is not defined for tables other than 2x2.
3) If you want to point to a site that is potentially useful,
you can do that without evoking a message from 1997.
The site nanmed above is sloppy, in my opinion, in the way
that it labels some of the screen. However, they are precise
enough that they start by naming the variation of Exact that
they are providing. (Freeman-Halton is not the only one possible.)
=== from the site
This unit will perform the Freeman-Halton extension of the Fisher
exact probability test for a two-rows by three-columns contingency
table, providing that the total size of the data set is no greater
than N=300. ...
===
--
Rich Ulrich