Is there such thing as a reverse T-Test?

Question

My understanding of the T-Test is that it is a test for checking if two samples are different enough to be considered from different distributions.

In the test the default position is that they are from the same distribution, and the burden of proof is showing that they are different.

Is there a reverse of this? The starting assumption is that the samples are from different distributions and I want to convince others that they are the same.

Many questions on site discuss the difficulty in swapping the null and alternative in hypothesis testing (though not all about the t-test). You can't show two things are the same, but if you define an equivalence (that population means within some specified distance of each other would be "equivalent") then you can get somewhere (this is called equivalence testing). — Glen_b, Oct 10 '17 at 07:49
One place to start --- https://stats.stackexchange.com/questions/21967/seeking-to-understand-asymmetry-in-hypothesis-testing — Glen_b, Oct 10 '17 at 08:49

score 3 · Answer 1 · answered Feb 25 '19 at 21:55

3

You can look into TOST (two one-sided t-tests). Essentially, they let you know whether 2 distributions are not significantly dissimilar.

answered Feb 25 '19 at 21:55

rahul bahadur

31

Is there such thing as a reverse T-Test?

1 Answers1