When you specify both BY and CLASS statements in PROC UNIVARIATE and any BY group has all missing values for the class variable(s), the procedure stop processings and issues the following message in the log:

WARNING: Insufficient nonmissing data.

Subsequent BY groups that have valid data are ignored.

Workaround

To circumvent this problem, specify the MISSING option in parentheses following the class variable(s). Undesired output for BY groups with missing class values will need to be manually deleted. Alternatively, the groups with missing data can be removed from the dataset before running the procedure.