topic Re: Conditional Individual Confidence Interval with Proc Mixed in Statistical Procedures

Conditional Individual Confidence Interval with Proc Mixed

AnastasiaK — Mon, 19 Dec 2016 16:34:04 GMT

I am running SAS proc mixed:

proc mixed data=Data;

class sample;

model result= time / DDFM=KENWARDROGER s cl E3 alphap=0.05 residual outp=mixedout;

random sample /s;

run;

I need to calculate Confidence Intervals for individual values for each sample separately.

We can call these intervals Prediction intervals.

With outp= I can obtain only CI's for mean values.

Can somebody help me?

Thank you in advance,

Anastasia.

Re: Conditional Individual Confidence Interval with Proc Mixed

SteveDenham — Tue, 27 Dec 2016 19:26:01 GMT

You can get EBLUPs (empirical best linear unbiased predictors) from the solution option on your RANDOM statement. You could add a CL option to get confidence bounds. The outp= only incorporates EBLUPs for repeated factors.

Steve Denham

Re: Conditional Individual Confidence Interval with Proc Mixed

AnastasiaK — Wed, 11 Jan 2017 09:18:54 GMT

Dear Steve,

thank you for you interest to my question.

I think that I didn't explained well my question.

I can have CIs from outp= .

But my problem is that I can not have Tolerance Intervals for 'result' variable. I need to estimate the intervals where individual measurements could be found if the experiment would be repeated again.

Thank you in advance,

Anastasia.

Re: Conditional Individual Confidence Interval with Proc Mixed

SteveDenham — Mon, 16 Jan 2017 20:13:41 GMT

I fear I am missing something. So you want confidence bounds on 'result' at various times for future repetitions at the same random effects? What about shifting to PROC GLIMMIX, where you can use the OUTPUT statement--

output out=out1 pred(blup ilink)=predicted lcl(blup ilink)=lower ucl(blup ilink)=upper;

This should give the blup's and confidence intervals on the original scale.

Steve Denham