ABSTRACT
This chapter showcases features of PROC SURVEYFREQ to facilitate categorical variables analyses. Whereas PROC SURVEYMEANS demonstrated in the previous chapter is the complex survey companion procedure to PROC MEANS, PROC SURVEYFREQ is the companion procedure to PROC FREQ. The chapter begins with a simple example pointing out how PROC SURVEYMEANS and PROC SURVEYFREQ produce equivalent results for certain univariate, descriptive statistics. The only difference is one of scaling: PROC SURVEYMEANS reports in terms of proportions, while PROC SURVEYFREQ reports in terms of percentages. We then segue into bivariate analyses and complex survey data adaptations to certain popular tests of association such as chi-square and likelihood ratio tests. Following that is a section devoted to odds ratios and a class of “risk” statistics that are useful for data that can be summarized in a 2 × 2 table. The concluding section illustrates multivariate capabilities of PROC SURVEYFREQ, albeit briefly, because most researchers prefer utilizing modeling procedures such as PROC SURVEYREG, PROC SURVEYLOGISTIC, and PROC SURVEYPHREG for multivariate analyses. Those three modeling procedures are covered in Chapters 5 through 7.
