topic Equivalence Test for Means: Adjust Confidence Intervals for the Differences? in Discussions

Equivalence Test for Means: Adjust Confidence Intervals for the Differences?

CoxPanther919 — Sun, 11 Jun 2023 11:30:53 GMT

Hello together,

I'm new to jmp17 and tried to do a TOST Test.

In Fit Y by X, I selected Equivalence Test --> Means.

Regarding the Forest Plot I wondered if the 90% Confidence Intervals for the Difference is the error bar in the plot and if there is a way to adjust the confidence interval?

Thanks!

Re: Equivalence Test for Means: Adjust Confidence Intervals for the Differences?

David_Burnham — Mon, 27 Mar 2023 15:15:11 GMT

In the dialog, when you enter the "difference" value, you can also enter the alpha-level. This will control the confidence interval width.

Re: Equivalence Test for Means: Adjust Confidence Intervals for the Differences?

Victor_G — Mon, 27 Mar 2023 15:55:51 GMT

Hi @CoxPanther919,

Welcome in the Community ! :)

Yes, the error bar represents the 90% confidence interval of the TOST you have realized.

Since you're doing a Two One-Sided Test, each one-sided test will use the alpha you have specified in the launch window (0,05 by default, so confidence level = 0,95), but you're doing two comparisons in a TOST : one for superior value, one for inferior value, so your "global" confidence for the TOST is 0,95 to the power of 2 (2 tests or comparisons done) = 0,90.

If you would like to have a 95% interval confidence, you will have to specify an alpha value of 0,025 (or say otherwise, confidence level of 0,975 for each test) : 0,975^2 = 0,95.

One practical example here :

Result :

I hope I understood your question and that this answer will help you,

Re: Equivalence Test for Means: Adjust Confidence Intervals for the Differences?

CoxPanther919 — Tue, 28 Mar 2023 12:51:57 GMT

Hi @Victor_G ,

thank you very much, this was exactly my question!

Best regards