How to prepare a data for Cochran-Mantel-Haenszel Test ?

I just think there are two few cases in the original dt after running the S/O, which worked and only apparent difference was the number of cases. Too many empties