topic Re: CALCULATE THE POWER by use of PROC SEQDESIGN in Statistical Procedures

CALCULATE THE POWER by use of PROC SEQDESIGN

Jack2012 — Thu, 05 Nov 2020 05:25:01 GMT

Hi All,

I am wondering if PROC SEQDESIGN could be used to calculate POWER of another alternative H2(vs H0) with calculated size by this procedure under specified alternative, like H1 (vs H0)?

It is quite common case to plan an IA under one assumption, like for survival HR1=0.7, and then you can calculate out the number of events.

If I want to know with such number of events, what is the power will be under the alternative is HR=0.6 ?

Thank you.

Re: CALCULATE THE POWER by use of PROC SEQDESIGN

Jack2012 — Thu, 12 Nov 2020 03:51:50 GMT

Can anybody assist me on this question?

Re: CALCULATE THE POWER by use of PROC SEQDESIGN

StatDave — Thu, 12 Nov 2020 14:56:40 GMT

Solving for power is not something SEQDESIGN does. However, with data from an interim look, you can use the CONDPOWER in PROC SEQTEST to show the conditional power, which is the conditional probability to reject the null hypothesis based on the observed statistic under a specific hypothesis. You can also use the PREDPOWER option in SEQTEST to display the predictive power, which is the posterior probability that the test statistic at the final stage would exceed the rejection critical value given the observed statistic. See "Stochastic Curtailment" in the Details section of the SEQTEST documentation for more information.

Re: CALCULATE THE POWER by use of PROC SEQDESIGN

Jack2012 — Thu, 19 Nov 2020 03:29:05 GMT

Great thanks Dave.

I see now.

Actually, we know when planning an IA, then PROC SEQDESIGN could be used to calculate the timepoints to the IA and the corresponding time to do the IA with a prespecified significance level, and targeted power with SAMPLESIZE statement.

Intuitively, I thought, how to calcute the power with prespecified timepoints, number of events/information with this.

Last but not least, I know this case now. Thank you.

Best regards,