topic Re: How do I get the Power printed out in SEQDESIGN in Statistical Procedures

How do I get the Power printed out in SEQDESIGN

normanbg — Tue, 29 Dec 2020 14:23:02 GMT

I am trying to run the following code for the SEQDESIGN procedure:

proc seqdesign altref=0.20
plots=boundary(hscale=samplesize)
plots=power
PSS(cref=0 0.5 1.0 1.5)
boundaryscale=pvalue;
OneSidedOBrienFleming: design nstages=2
method=obf
alt=upper stop=both
alpha=0.05 beta=0.20;
samplesize model(ceiladjdesign=include)
=onesamplefreq(nullprop=0.75 prop=0.95 ref=prop)
inputnobs(sample=one N=38);
ods output AdjustedBoundary=Bnd_Prop;
run;

but I can't figure out how to get the power (1-beta) at each stage for both the null hypothesis and the alternative (ie, the power to reject H0 when it is indeed false and the power to reject H1 when H1 is indeed false). It must be there somewhere, I just can't find it.

Any help would be greatly appreciated.

Norman

Re: How do I get the Power printed out in SEQDESIGN

StatDave — Tue, 29 Dec 2020 15:57:00 GMT

Try the ERRSPEND option in the PROC statement

Re: How do I get the Power printed out in SEQDESIGN

normanbg — Tue, 29 Dec 2020 16:43:51 GMT

I tried. This is what I got. I couldn't save the screen shot in an acceptable format so I put 3 copies of it in a zip file in the hope that you will be able to read one of them.

Are these 4 entries the 4 power values I am looking for? If so, which is what?

Thank you for trying to help.

Re: How do I get the Power printed out in SEQDESIGN

StatDave — Tue, 29 Dec 2020 17:09:46 GMT

For your two-stage design, the beta values at each stage is as shown: .09 and .2. For info on this, see the "Details: Error Spending Methods" section in the PROC SEQDESIGN documentation.

Re: How do I get the Power printed out in SEQDESIGN

normanbg — Tue, 29 Dec 2020 21:31:21 GMT

I still don't get it. It's not surprising that the power (to reject a false H0) at stage 2 is 80%, because that was part of the initial design. But a power at stage 1 of 91% when the sample size there is half that at stage 2—I think not. What am I missing?

Thank you for your perseverance.

Re: How do I get the Power printed out in SEQDESIGN

StatDave — Wed, 30 Dec 2020 15:55:00 GMT

This might be clearer if you use the STOPPROB option instead. See the description of the values in the resulting table at the end of the "Details: Table Output" section of the SEQDESIGN documentation. You will see that the probability of stopping the trial for acceptance under the alternative (CRef=1) at stage 1 is .09 and .2 at stage 2. The probability of rejecting is .31 at stage 1 and .8 at stage 2. The overall probability of stopping the trial early at stage 1 is .4.

For multi-stage designs, once you start collecting data and doing an interim analysis, you might also be interested in conditional and predictive power, which are both available with options in PROC SEQTEST.

Re: How do I get the Power printed out in SEQDESIGN

normanbg — Wed, 30 Dec 2020 16:42:03 GMT

Bingo! Just what I was looking for. Thank you, thank you very much.