To do this, you treat the complete data set as the population and the proportion(s) from it as fixed values. For example, if the comparison is with respect to a binary variable (with values such as yes or no), then this means you want to test that the proportion responding yes in the sample equals the known proportion responding yes in the population (complete data set).

You can use the BINARY(P=p) option or the TESTP= option in TABLES statement of PROC FREQ to do this. For a binary variable, specify the population (complete data set) proportion in the BINARY(P=p) option. For a multinomial variable, specify the population proportions in the TESTP= option. PROC FREQ will compute a chi-square test that the sample proportions equal the population (complete data set) proportions.

For example, if the variable of interest is RESPONSE with possible values 1, 2, and 3 and the proportions in the population or complete data set are .25, .50, and .25, respectively, then the following statements produce the test:

proc freq;
  tables response / testp=(.25 .50 .25);
  run;

Operating System and Release Information