Paired-samples test of equivalence

[#1]

Hi,

Is it possible within JMP Pro to perform a paired-samples test of equivalence? I can only see an option under the Fit Y by X platform to carry out an equivalence test without a grouping variable.

Thanks in advance

[#2 · Accepted solution]

You generally use the Analyze > Specialized Modeling > Matched Pairs for this purpose, but this platform does not currently provide an equivalence test. For now, use a formula to compute the difference in the data table. I mocked up paired responses Y1 and Y2.

Select Analyze > Distribution and launch it with Difference in the Y role. Click the red triangle next to Difference and select Test Equivalence:

Specify 0 for the expected mean and whatever limit you use for practical equivalence. I claim that my example is equivalent if the result is within 1 unit:

Click OK, and you will get the result:

This example shows that these two responses are practically equivalent because the two-sided tests are significant and the confidence interval estimate of the mean is entirely within the +/- 1 interval specified.

[#3]

Thank you so much, that's so helpful!

[#4]

Hi Mark,

Is this still the appropriate method for doing a paired TOST or is there new capability in JMP 17?

Thank you,

Stephanie

[#5]

Yes, this way is still the most appropriate method using JMP 17..

Equivalence testing has been added to Matched Pairs in JMP 18, which will be out in a couple of months. You won't have to calculate the difference prior to launching the platform. There is no preparation of the data required. Also, it provides a richer set of tests and a non-parametric form. The reporting is richer, too, consistent with recent additions of equivalence tests in other platforms.