CMH weighted Miettinen and Nurminen CI

[dcebrik]

Is there a SAS procedure to calculate Cochran Mantel Haenszel weighted Miettinen and Nurminen confidence intervals for risk difference in stratified samples?

[SAS_Rob]

Yes, with the RISKDIFF(COMMON) option on the TABLES statement you should get Miettinen-Nurminen (score) confidence limits.

http://go.documentation.sas.com/?docsetId=statug&docsetTarget=statug_freq_details62.htm&docsetVersio... 

[dcebrik]

Thank you for your time, however this method doesn't appear to use CMH weights.

[PJL]

Hi,

Yes, it looks like it also uses a different method of combining the per-stratum estimates, from Agresti, rather than the stratified calculations described in the Miettinen & Nurminen paper.

(With apologies to the moderator if this is inappropriate) if you are able to use R, CMH-weighted stratified MN confidence intervals can be obtained using the 'ratesci' package. This also includes a skewness-corrected version that results in more evenly-distributed tail probabilities.