MISSING Subcommand (CSCOXREG command)

The MISSING subcommand specifies how missing values are handled.

• In general, cases must have valid data for all design variables as well as for the dependent variable and any covariates. Cases with invalid data for any of these variables are excluded from the analysis.
• There is one important exception to the preceding rule. This exception applies when an inclusion probability or population size variable is defined in an analysis plan file. Within a stratum at a given stage, if the inclusion probability or population size values are unequal across cases or missing for a case, then the first valid value found within that stratum is used as the value for the stratum. If strata are not defined, then the first valid value found in the sample is used. If the inclusion probability or population size values are missing for all cases within a stratum (or within the sample if strata are not defined) at a given stage, then an error message is issued.
• The CLASSMISSING keyword specifies whether user-missing values are treated as valid. This specification is applied to categorical design variables (that is, strata, cluster, and subpopulation variables), the dependent variable, and any factors.

EXCLUDE. Exclude user-missing values among the strata, cluster, subpopulation, dependent variable, and factor variables. This is the default.

INCLUDE. Include user-missing values among the strata, cluster, subpopulation, dependent variable, and factor variables. Treat user-missing values for these variables as valid data.