topic Re: Resetting Lower Confidence Limit (LCL) when negative LCL arises from SAS Survey Procedure in Statistical Procedures

Resetting Lower Confidence Limit (LCL) when negative LCL arises from SAS Survey Procedure

SAS_User8 — Fri, 04 May 2018 14:53:14 GMT

Is it possible to add an option to the SAS Survey Commands to reset negative LCL to zero?

Re: Resetting Lower Confidence Limit (LCL) when negative LCL arises from SAS Survey Procedure

ballardw — Fri, 04 May 2018 16:46:16 GMT

Send your output to a data set and use data step program to reset such things. You might want to look at upper bounds as well if your confidence limits are on percentages as they can exceed 100 percent mathematically if not practically.

Re: Resetting Lower Confidence Limit (LCL) when negative LCL arises from SAS Survey Procedure

SAS_User8 — Fri, 04 May 2018 16:50:41 GMT

Thanks! Yes, that's what we do, but I was wondering if SAS could introduce an option in the SAS survey commands to request that change be made automatically. It's a bit tedious to do manually in data step.
Thanks for considering!
a

Re: Resetting Lower Confidence Limit (LCL) when negative LCL arises from SAS Survey Procedure

Rick_SAS — Fri, 04 May 2018 18:47:44 GMT

What parameter are you estimating? Post your code. There might be something we can suggest

Many SAS procedures ensure that confidence intervals make sense for the parameter. For example, if you use PROC FREQ to estimate a binomial proportion, the CI for the proportion is always a subinterval of [0,1]. Similarly, a confidence interval for a variance or an odds ratio will never include a negative value.

If your statistic is assumed to be normally distributed, SAS will carry out the usual computations. SAS does not know when a parameter is bounded unless you tell it.