1

We have two samples with different sizes: 42 and 47 and we have to verify if the means of the two samples are equals. For verify this with t-test, we should verify first if the variances are different or equal with F-test?

Math Info
  • 121
  • 5
  • 1
    Better to use the [Welch-type version of the t-test](https://en.wikipedia.org/wiki/Welch%27s_t-test) than use an F-test on the variance. However those sample sizes are reasonably close so moderate differences in the relative variances shouldn't be all that much of a problem. I'm pretty sure this has been answered before. – Glen_b May 10 '17 at 12:41
  • @Glen_b , when I use F-test the null hypothesis is rejected. – Math Info May 10 '17 at 12:48
  • How does that change the advice to do something else without even looking at the F-test? The problem with what you're doing arises *because* you will change your actions with the t-test based on looking at that F-test. Better not to look at all; if you're not comfortable that the t won't be badly affected you should have just done the Welch test in the first place. – Glen_b May 10 '17 at 12:55
  • Besides the indicated post this was closed as a duplicate of, also see the discussion inside the relevant section of the answer [here](https://stats.stackexchange.com/questions/121852/how-to-choose-between-t-test-or-non-parametric-test-e-g-wilcoxon-in-small-sampl/123389#123389), and the extended discussion [here](https://stats.stackexchange.com/questions/87215/does-a-big-difference-in-sample-sizes-together-with-a-difference-in-variances-ma) – Glen_b May 10 '17 at 13:07

0 Answers0