2

I have read several posts about equivalence testing, I can also see that there are many questions about it on CV... but I feel this type of tests might be tricky, especially for people like me with little statistician background. To reduce the scope of my question, let's assume that I am interested in sample comparison where a traditional t-test can apply (continuous data from a simple random sample, etc.).

Apparently, when we are talking about equivalence testing, the standard approach seems to be the following (which seems less obvious to manipulate). Then I read about TOST which seems very similar actually to the previous method.

I have seen also that people are talking about variance test and define an acceptable value for variance/standard deviation that will prove that the sample are similar. I guess confidence intervals might also be used to interpret equivalence...

So, what are the options when we want to test equivalence?

Alexis
  • 26,219
  • 5
  • 78
  • 131
l.apprenti
  • 29
  • 7
  • 1
    Compare your parameter of interest. [Here is an example on my former blog](https://stla.github.io/stlapblog/posts/EquivalenceExceedanceProbabilities.html), where I do an equivalence test for exceedance probabilities. – Stéphane Laurent Sep 17 '21 at 11:33
  • Thanks @StéphaneLaurent, seems very interesting! I need to dive into it now... – l.apprenti Sep 17 '21 at 12:28

0 Answers0