CMH Analysis - Issue with Results

[mattkumar]

Hello,

We've encountered what seems to be an issue regarding the output of the CMH function when specifying the TYPES argument.

For reference, we've also used SAS 9.4 to do the similar analysis. In short, it seems using the TYPES argument results in mismatching of the degrees of freedom for each hypothesis test, leading to incorrect P-Values.

We've created a reproducible example that can be accessed here: https://github.com/phuse-org/CSRMLW/blob/main/cmh/Docu/final_report/vcdExtra_issue.Rmd

Any thoughts around this would be greatly appreciated. Many thanks for your work on the vcd and vcdExtra packages!

Best,
Matt

[friendly]

@mattkumar thank you for this detailed analysis and reproducible example. I'm checking out ShuguangSun's proposed fix.