When you specify a SIZE variable, any sampling units that have missing or nonpositive size measures are excluded from the sample selection. The procedure provides a log note that reports the number of observations omitted due to missing or nonpositive size measures.
If you do not use a SAMPLINGUNIT statement with the SIZE statement, your sampling units are input data set observations, and observations that have missing or nonpositive size measures are excluded from the sample selection. If you do use a SAMPLINGUNIT statement with the SIZE statement, the procedure computes sampling unit size by summing the size measures of all observations in the unit. When summing the observation size measures, the procedure omits any observations that have missing or nonpositive size measures. If the size of an entire sampling unit is missing or nonpositive, the procedure excludes that unit from the sample selection. When a sampling unit is selected, the output data set includes all observations that belong to the selected unit, regardless of whether an observation’s size measure is missing.
If you provide stratum-level design or allocation information in a secondary input data set, the variable values should be nonmissing. For example, if a stratum value of _NSIZE_ (or SampleSize) in the SAMPSIZE= secondary input data set is missing or negative, PROC SURVEYSELECT cannot select a sample from the stratum. The procedure gives an error message and skips the stratum. Similarly, if other secondary data set variables have missing values for a stratum, a sample cannot be selected from the stratum. These variables include _NRATE_, _MINSIZE_, _MAXSIZE_, _CERTSIZE_, and _CERTP_. Additionally, if any of the sample allocation variables in the secondary input data set have missing or nonpositive values, PROC SURVEYSELECT cannot compute the sample allocation. Variables that provide information for allocation include _ALLOC_, _VAR_, and _COST_. See the section Secondary Input Data Set for details.