The CATMOD Procedure

Missing Values

Observations with missing values for any variable listed in the MODEL or POPULATION statement are omitted from the analysis.

If the WEIGHT variable for an observation has a missing value, the observation is by default omitted from the analysis. You can modify this behavior by specifying the MISSING= option in the MODEL statement. The option MISSING=value sets all missing weights to value and all missing cells to value. The option MISSING=SAMPLING causes all missing cells in a contingency table to be treated as sampling zeros.

Any observation with nonpositive weight is also, by default, omitted from the analysis. If it has zero weight, then you can specify the ZERO= option in the MODEL statement.