Re: Comparing two cross product tables

On 23 Oct 2005 09:23:04 -0700, schwarzgroup@xxxxxxxxxxx wrote:

> Hi there-
>
> I'm doing a study and I have two different cross product tables that
> are both 2x2. Basically, I want to see if the tables are different
> from either other statistically. I know that when looking at just one
> table, you can do a chi-square test on the diagonals to see if it's
> different than 50/50. I don't care about that really, just the
> differences between the two 2x2 tables as I've separated data based on
> a characteristic.

Question about terminology:
Are "cross product tables" what I would call 2x2
contingency tables? Further detail you provide makes
me wonder more --

The usual test on a contingency table compares the
cells to what is expected from the marginals, and I
would not call that a "test on the diagonals." A test
that compares two tables (that are 2x2 contingency
tables) is usually a test that compares the two odds-ratios.

Another test on a 2x2 table is a test-on-changes, the
McNemar test, which uses the OFF-diagonals, and is
equivalent to a sign-test (50/50). It would take careful
attention to details to set up a proper hypothesis
between two of those tables.

--
Rich Ulrich, wpilib@xxxxxxxx
http://www.pitt.edu/~wpilib/index.html
.

Loading