04-29-2015 02:10 AM
When using survey data (with clustering and stratification), what are the recommended approaches to estimate a statistic which is not estimated by the SAS survey procs? For example, a confidence interval for the ratio of the mean of variable 1 to the standard deviation of variable 2. Can the survey procs be tweaked for this? Any pointers to a replication (bootstrap or jackknife) approach?