Calculating confidence intervals for proportion when the sample size is small

[Emma_at_SAS]

Hello,

I have some samples from a binary distribution and I want to estimate confidence intervals for the proportion of success, however, my samples are small. For example, in a sample of 6, I have 4 successful observations. The other note is that this is a subsample from a survey and I need to use a SURVEY procedure. I appreciate your thoughts and help on this.

Thanks

[Rick_SAS]

I suggest you look at the section "Replication Variance Estimation" in the SURVEYFREQ documentation.  Without knowing more about your survey design, I would suggest that you consider using bootstrap estimates for the CI for proportion: VARMETHOD=BOOTSTRAP. The documentation states the requirements for using the bootstrap and also provides other alternatives, such as the Balanced Repeated Replication (BRR) method.