How to perform CMH test in survey data using Stata?

[Qiguo Lian]

Besides, is there anyone knows the differences between -emh- and -opartchi-? Seems that they can do the same thing but the results are not exactly same.
Thanks in advance!

[Brendan Halpin]

Cochrane-Mantel-Haenszel? By survey data, do you mean complex sampling structure?

[Steve Samuels]

The FAQ warn against asking two different questions, one not a natural extension of the other, because answers to each will confuse the thread. Ask the second question separately.


svy: clogit is the survey alternative to the Mantel-Haenszel test. Every member of a matched set must have the same survey weight. If that's not the case, then give all the controls of each case (1:n matching) the case's sampling weight.

Brendan, the "C" in "CMH" is William Cochran, who published a version in 1954. The MH innovation was to consider both row and column totals in each 2 x 2 table as fixed, which led to a finite population correction of the variance. If I remember correctly, the classic Snedecor and Cochran text "Statistical Methods" presented the C and MH versions in different sections, perhaps in different chapters.

References:

Cochran, WG. 1954. Some methods for strengthening the common chi square tests. Biometrics 10, no. 4: 417-451.

Mantel,.N., and Haenszel, W., "Statistical aspects of the analysis of data from retrospective studies of disease," Journal of the National Cancer Institute, 22 (1959), 719-48.

[Qiguo Lian]

The FAQ warn against asking two different questions, one not a natural extension of the other, because answers to each will confuse the thread. Ask the second question separately.


svy: clogit is the survey alternative to the Mantel-Haenszel test. Every member of a matched set must have the same survey weight. If that's not the case, then give all the controls of each case (1:n matching) the case's sampling weight.

Brendan, the "C" in "CMH" is William Cochran, who published a version in 1954. The MH innovation was to consider both row and column totals in each 2 x 2 table as fixed, which led to a finite population correction of the variance. If I remember correctly, the classic Snedecor and Cochran text "Statistical Methods" presented the C and MH versions in different sections, perhaps in different chapters.

References:

Cochran, WG. 1954. Some methods for strengthening the common chi square tests. Biometrics 10, no. 4: 417-451.

Mantel,.N., and Haenszel, W., "Statistical aspects of the analysis of data from retrospective studies of disease," Journal of the National Cancer Institute, 22 (1959), 719-48.

Thanks, Steve.