If BY processing is used along with a WEIGHT or FREQ statement, and the weights or frequencies are all missing, zero or negative in the first BY group, then PROC UNIVARIATE will fail to generate any results for subsequent BY groups even if they have valid observations. The following incorrect Note appears in the log:

  NOTE: There are no observations with valid WEIGHT/FREQ values in the input data set.

To circumvent this problem, take measures to assure that the first BY group contains valid observations. For example, if the first BY group has all missing, zero or negative weight/freq values, omit it from the analysis using a WHERE clause.