I answered these homework questions but I was told that at least one of my answers is wrong. However, I can't tell which one I answered wrong. What incorrect assumptions have I made?


enter image description here


3a) Welch's T-test would be most appropriate since we have normal data with unequal variances.

3b) A Wilcoxon-Mann Whitney U Test appears to be the best option since transformations have failed and we must now use a non-parametric test for these 2 independent samples.

3c) The data has unequal variances and is not normal so we should also use a Wilcoxon-Mann Whitney U Test here too.

  • $\begingroup$ In (c), consider the interpretation of "p-values > $\alpha$" (not that I agree with the reasoning behind these questions). $\endgroup$ – mark999 Nov 16 '16 at 5:23
  • $\begingroup$ I believe I did. When p > α in Levene's test, there are unequal variances. When p > α in the Shapiro-Wilk test, the distribution is not normal. Or am I incorrect? $\endgroup$ – orangebull Nov 16 '16 at 5:35
  • $\begingroup$ en.wikipedia.org/wiki/Levene%27s_test $\endgroup$ – mark999 Nov 16 '16 at 5:39
  • $\begingroup$ If they really will not tell you which one is wrong you might consider changing course and asking for your money back. $\endgroup$ – mdewey Nov 16 '16 at 12:25

I have a number of issues with this question, but leaving those aside, it's clear that your answer to (c) is not what they're seeking.

When the p-value is large the null should not be rejected. In significance testing, the null hypothesis is rejected when the p-value is smaller than the significance level ($\alpha$).

It would probably be a good idea to review your understanding of the basics of hypothesis testing.

| cite | improve this answer | |
  • $\begingroup$ I just noticed my mistakes. The variances are equal and the distribution is normal. But does that mean that an independent t-test would work best or would a correlation/regression test be more appropriate since we are working with 2 quantitative variables? $\endgroup$ – orangebull Nov 16 '16 at 17:16
  • $\begingroup$ Failing to reject the nulls doesn't tell us that the variances are actually equal or that the distribution is actually normal. You're working with the variables wing-length and species; you can do equivalent analyses either with a t-test or regression. $\endgroup$ – Glen_b Nov 16 '16 at 23:00

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.