topic Re: Using PROC SURVEYFREQ instead of PROC FREQ for 95% CI? in Statistical Procedures

Using PROC SURVEYFREQ instead of PROC FREQ for 95% CI?

KnicksFan11 — Thu, 10 Mar 2022 22:04:56 GMT

Hi,

I was wondering if I can use PROC SURVEYFREQ instead of PROC FREQ to get 95% CI for simple proportions in a 2 x 2 categorical table. It is cumbersome to use the binomial function in PROC FREQ to get confidence limits and involves a lot of data manipulation. It is much easier to get rowpcts in SURVEYFREQ. The Chi-Sq values seem to line up, however, the CI between the procedures are very slightly different. I am not using complex survey data, but does SURVEYFREQ work like regular old FREQ when it is fed a simple data without weights, clusters, or strata?

Thanks!!

Re: Using PROC SURVEYFREQ instead of PROC FREQ for 95% CI?

sbxkoenk — Fri, 11 Mar 2022 08:50:16 GMT

See here :

SAS® 9.4 and SAS® Viya® 3.5 Programming Documentation | SAS 9.4 / Viya 3.5
SAS/STAT User's Guide
The SURVEYFREQ Procedure
Confidence Limits for Proportions
https://go.documentation.sas.com/doc/en/pgmsascdc/9.4_3.5/statug/statug_surveyfreq_details41.htm?homeOnFail

Koen

Re: Using PROC SURVEYFREQ instead of PROC FREQ for 95% CI?

SAS_Rob — Fri, 11 Mar 2022 14:57:49 GMT

If you want row percentages and confidence limits from Proc FREQ in a 2x2 table setting then you could use the RISKDIFF option on the TABLES statement.

Using SURVEYFREQ is also an option, but the method by which it computes standard errors might lead to slightly different answers than you would get using Proc FREQ, even if you do not have any design variables.