I had a prof once refer to doing science in excel as doing watch repair with oven mitts. Nevertheless, I'm trying...

I'm trying to do a TOST test using excel. I'm basing the equivalence of variances on the excel F-test. If I run both t-tests with a hypothesized mean difference of $+\Delta$ and then $-\Delta$, with $\alpha$ = 0.05, is my final, combined $\alpha$ 0.10?

To elaborate: I made three measurements before and after an event. I hope to claim statistically that the event had no effect. I have a measurement uncertainty (0.14) for each measurement, and six values (5.7, 5.8, 5.5 before; 5.6, 5.4, 5.6 after). Because 0.1 is my measurement uncertainty, I thought that it would be a reasonable choice for $\Delta$.

For the excel t-test calculator, I provide the values, an anticipated difference in mean (0.14) and $\alpha$, and run the calculator twice with the before and after swapped (because excel won't take a negative anticipated difference). I get two t-stats to which to compare to the t critical one tail value. One of these t Stats is -2.47 which is less than -2.13, the t Critical one tail. Is this sufficient to claim equivalence per TOST? What is my final $\alpha$ if I ran both of these t-tests with $\alpha$=0.05?

  • $\begingroup$ FWIW, note that TOST stands for "Two One-Sided T-tests". Since there is an Excel function for a t-test, you can do this yourself pretty easily. $\endgroup$ – gung - Reinstate Monica Jun 13 '19 at 19:59
  • $\begingroup$ thanks for the quick response. If I'm understanding correctly (and borrowing heavily from the cross validated page on TOST), I run t-tests in excel hypothesizing a difference of +/- delta, where delta is my chosen limit in the same unit as the mean, I get a t stat to which I can compare to the 't critical one-tail'. I'm not sure I follow the description of alpha: if I run both of these tests with alpha = 0.05, is my overall alpha 0.1? $\endgroup$ – Nick Jun 13 '19 at 20:19
  • $\begingroup$ I added more info, above. I also found this page, if anyone follows this thread looking for the same thing: daniellakens.blogspot.com/2016/12/… (it says R, but it has an excel spreadsheet, as well). $\endgroup$ – Nick Jun 13 '19 at 21:55
  • $\begingroup$ Nick, your overall $\alpha$ does not sum the $\alpha$s of these two tests because their $H_{0}$s are non-intersecting. $\endgroup$ – Alexis Nov 27 '19 at 6:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.