topic Calculating 90% CI for specified percentile values in SAS Programming

Calculating 90% CI for specified percentile values

cloverhorizon — Thu, 24 Feb 2022 16:18:24 GMT

I am trying to calculate a 90% confidence interval in proc univariate for the 2.5th and 97.5th percentiles. I added in pctlpts to calculate the percentiles as they are not in the default values, and the output gives me the values. After adding in ciquantdf to get the 90% CI's of these percentiles though, the output no longer lists these specified percentiles, only the defaults. Does anyone know how to calculate the 90% CI's for these specified percentiles, even if not in univariate? Thank you!

proc univariate data=itd ciquantdf(alpha=.1);
var res;
output out=per pctlpts = 97.5 2.5 pctlpre = P_;
run;

Re: Calculating 90% CI for specified percentile values

sbxkoenk — Thu, 24 Feb 2022 17:51:23 GMT

Hello,

Have a look at this :

Compute confidence intervals for percentiles in SAS
By Rick Wicklin on The DO Loop May 6, 2013
https://blogs.sas.com/content/iml/2013/05/06/compute-confidence-intervals-for-percentiles-in-sas.html

Cheers,

Koen

Re: Calculating 90% CI for specified percentile values

ballardw — Thu, 24 Feb 2022 18:02:30 GMT

You would have to request those items in the output statement. Example (may not have values for your data though)

output out=per pctlpts = 97.5 2.5 pctlpre = P_    
    ciquantdf=(lowerpre=LCL UpperPre= UCL);

The statistic request is again Ciquantdf. The LowerPre and UpperPre options specify the prefix for the variable names like Pctlpre does. The number of values and ranges might mean that some of the intervals aren't calculated, just as in the table output.