### spss >> Mantel-Haenszel method for matched sets (case-controls studies)

How can I run the Mantel-Haenszel method for matched sets
(case-controls studies)in SPSS11.5 ???

### spss >> Mantel-Haenszel method for matched sets (case-controls studies)

It's a statistics option under Crosstabs. Put your explanatory
variable in the column, your response variable in the row,
and your stratification variable(s) in the layer(s).

e.g., in syntax:
crosstabs disease by smoke by AgeCat /statistc = cmh(1) /cells = count .
--
=-=-=-=-=-=-=-=-=-==-=-=-=
Mike Lacy, Ft Collins CO 80523
Clean out the 'junk' to email me.

### spss >> Mantel-Haenszel method for matched sets (case-controls studies)

Hi Mike. I think the original poster was asking about a
case-control study with *matched pairs* of cases and
controls. I don't see how your analysis takes into account
the matching.

I don't have it in front of me right now, but I think the
following chapter might have some revlevant general info
(i.e., not specific to SPSS):

Hennekens, C.H., & Buring, J.E. (1987). Analysis of
epidemiological studies: Evaluating the role of confounding.
In Hennekens, C.H., & Buring, J.E. (Eds.), "Epidemiology in
Medicine. Little, Brown & Co.

Cheers,
Bruce
--
Bruce Weaver
XXXX@XXXXX.COM
www.angelfire.com/wv/bwhomedir/

### spss >> Mantel-Haenszel method for matched sets (case-controls studies)

Duh...Bruce is right, I just spaced out the matched feature.
I was a bit confused, I guess, since I would describe something
as either an MH approach or a matched pair approach, but not
both phrases in the same sentence. (There is a connection, though,
see below.) In any event, presuming that what is wanted
is a matched pair analysis for a binary exposure variable,
I have two SPSS solutions:

1) If I recall correctly, if you use the matched pair as
the stratification variable (*lots* of strata), the Mantel-Haenzsel
analysis still works fine. I think the story is that it works well as
the sample size within strata gets large, or as the number
of strata gets large. I've never tried this in SPSS; I don't
know if it's smart enough to not get screwed up by the tied
pairs. Anyway, the SPSS approach would be to have the
case and control on different "cases" in SPSS, then do
a crosstabs of disease by exposure, using the pair ID
variable as the layer. Don't know if this will work,
would be curious to hear.

2)Another approach, now that I think of it, should work fine:
Start off by restructuring the data so that the case and
the control are on the same "case" in SPSS. Do a crosstab
of the exposure variable for the case vs. the exposure
variable for the control. Choose the McNemar chi-square,
which turns out to be identical to the MH test (see
Rothman and Greenland 1998, Modern Epi. 2nd Ed.)
From the table, the ratio of concordant to discordant
pairs will give an odds ratio estimate that is the
same as the MH estimate.

--
=-=-=-=-=-=-=-=-=-==-=-=-=
Mike Lacy, Ft Collins CO 80523
Clean out the 'junk' to email me.