With METHOD=SYS, SAMPRATE= and NMIN= specified on the PROC SURVEYSELECT statement, the NMIN= option may be ignored for some initial seed values when CEIL(SAMPRATE*N) is equal to the value specified for NMIN. N is the total number of observations in the stratum or data set. These is no way to predict which seed values will cause this to occur.

The problem does not occur with METHOD=SRS, USR or SEQ. To circumvent the problem when using METHOD=SYS, the affected strata will need to be extracted manually from the DATA= data set and merged with the sample selected by PROC SURVEYSELECT.