To go along with Tom's answer: Along with the custom missing value .A and such you can ( fairly highly recommended from my point of view) also assign custom formats to describe the specific from of missing. The following stub program reads in a question with 5 valid response categories and a missing (could be skip pattern somewhat similar to original question), assigns special missing to categories that will generally be ignored for MOST analysis, a format to display the meaning of the special missings, and an example with proc freq of analysis with and without missing values included. proc format library=work; invalue Q1Missing 7=.D 9=.R 1,2,3=_same_ other= . ; value q1missing .D ="Don't Know" .R ="Refused" 1 ="Yes" 2 ="No" 3 ="Not sure" . = "Not Answered" ; run; data example; input x q1missing.; format x q1missing.; datalines; 1 2 3 7 9 . ; run; Proc freq data=example; tables x / missing; tables x; run;
... View more