topic Re: PROC POWER : Alpha with TEST=Equiv_Ratio in Statistical Procedures

PROC POWER : Alpha with TEST=Equiv_Ratio

Clg — Mon, 15 Feb 2021 16:02:08 GMT

Hello everyone,

When I use the PROC POWER with PAIREDMEANS and TEST=EQUIV_RATIO, what Confidence Interval provides the procedure for the sample size with alpha=0.05 ? A 95% CI or a 90% CI because equiv_ratio is based on two one-side tests ?

Thank you,

Clg

Re: PROC POWER : Alpha with TEST=Equiv_Ratio

jiltao — Mon, 15 Feb 2021 21:51:14 GMT

For an equivalence test the specified ALPHA= value provides the sample size for a (1-2α)100% confidence interval on the mean ratio. So if you specify alpha=0.05, then you get a 90% confidence interval on the mean ratio.

Re: PROC POWER : Alpha with TEST=Equiv_Ratio

Clg — Mon, 22 Mar 2021 07:38:30 GMT

Hello,

I have another question. Now, if I use TEST=RATIO (and not equiv_ratio) and that I want an 90% confidence interval with a one-sided test; I use "Sides=U" (or L) and I put alpha=0.1, it's right ?

Thank you.

Re: PROC POWER : Alpha with TEST=Equiv_Ratio

jiltao — Mon, 22 Mar 2021 12:32:24 GMT

Yes, that is correct.

Re: PROC POWER : Alpha with TEST=Equiv_Ratio

Clg — Tue, 23 Mar 2021 07:40:53 GMT

My apologizes for this another question. If I done the following SAS Code, I compare if the ratio 1.5 is different that 1.25 with a 90% confidence interval also?

proc power;
pairedmeans test=ratio alpha=0.1
meanratio=1.5
nullratio=1.25
cv=0.17
npairs=.
corr=0
sides=2
power=0.9;
run;

Thank you

Re: PROC POWER : Alpha with TEST=Equiv_Ratio

jiltao — Tue, 23 Mar 2021 14:05:23 GMT

It is basically correct.