The population we want to use must be age 55 and over and are linked to a provider. Unfortunately, we only have 14,400 patients that meet the criteria. I re-ran the program with just Age and Sex as my strata per your suggestion and still got the same error. I then took out Sex and that just left me with Age as my strata and still got the same exact error. I'm a beginner at this and would appreciate any other suggestions.

Thanks.

--- "David L. Cassell" <cassell.david@EPAMAIL.EPA.GOV> wrote: > s huynh <senthuynh@YAHOO.COM> wrote: > > This is my first time with simple random sampling. > I > > got the following error. What does this mean and > how > > do I get around this error? Thanks. Sen > > > > > > > > 47365 PROC SURVEYSELECT DATA=FINALPOP > > OUT=SELMRNS > > 47366 SEED=78463905 > > > > 47367 METHOD=SRS > > > > 47368 SAMPSIZE=9; > > > > __________ > > > > __________ > > > > __________ > > > > 1 > > > > 1 > > > > > > 47369 STRATA PCP_NAME AGE PAT_SEX; > > > > 47370 > > RUN; > > > > ERROR: The sample size, 9, is larger than the > number > > of sampling units, 7. > > ERROR: The sample size, 9, is larger than the > number > > of sampling units, 7. > > ERROR: The sample size, 9, is larger than the > number > > of sampling units, 7. > > NOTE: The above message was for the following > stratum: > > > > PCP_NAME=Bob Mcdonald AGE=55 PAT_SEX=F. > > > > ERROR: The sample size, 9, is larger than the > number > > of sampling units, 5. > > ERROR: The sample size, 9, is larger than the > number > > of sampling units, 5. > > ERROR: The sample size, 9, is larger than the > number > > of sampling units, 5. > > NOTE: The above message was for the following > stratum: > > > First, this is NOT simple random sampling, even if > the > METHOD=SRS part may mislead you. This is stratified > sampling > with a multi-level system for strata. > > And, unbelievably (for those with lots of experience > with SAS), the > error message is interpretable. It is warning you > of exactly the > problem I forecast in my earlier message. I > predicted that you might > not be able to get your sample 'smoothed out' as > much as you wanted, > simply because some of your stratum levels are so > darn tiny. I > hate it when I'm this accurate. > > Now you have to decide how *you* want to handle > this. You can't > have it both ways. Either you'll have to change how > you stratify > everything (and make it into fewer stratum levels), > or you'll have > to go with an approach that will not give you as > even a count per > stratum level. > > Why do you only have 14000 records to start with? > Why do you need to spread things so evenly across > the values of > PROVIDER? > Why can't you just go with levels of age and sex as > you strata, > and not worry so much about the providers? > > Think things over... > David > -- > David Cassell, CSC > Cassell.David@epa.gov > Senior computing specialist > mathematical statistician >

__________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com