Confidence intervals around weighted proportions

NG1 — Thu, 17 Jul 2014 20:46:25 GMT

I conducted a survey in multiple healthcare clinics to assess the prevalence of certain conditions (e.g., diabetes). I am trying to calculate the proportion and 95% CI for those with the condition after weighting by size of clinic (i.e., number of total clinic visits that year).

I tried using the following procedure:

proc surveyfreq data=survey;

tables diabetes / cl;

weight visits;

run;

Question --- I want to calculate a 95% CI that takes into account the variance from each stratum (each facility) that was used to calculate that weighted proportion. Does this code do this? Or does it merely take the final weighted proportion and then calculate a 95% CI around it assuming n observations? Also, I saw on a related post something about taking target population into account when calculating confidence limits using proc surveyfreq. Is there a way to do this?

Appreciate your help!

Re: Confidence intervals around weighted proportions

ballardw — Thu, 17 Jul 2014 22:09:50 GMT

If you sampled by strata then you definately need a STRATA statement and possible associated sample rates with the RATE= option or different counts using a Total=option dataset. Look in the Surveyfreq online help for the descriptions of the data sets required with the Rate or Total option.

topic Re: Confidence intervals around weighted proportions in SAS Procedures

Confidence intervals around weighted proportions

Re: Confidence intervals around weighted proportions