0

I am using Stata for my analysis. I am analysing a numerical variable over a categorical one (with 3 groups). After confirming they are normally distributed, I did a one way ANOVA test. But Bartlett's test showed me that the variances are not equal. What should I do next? Do a Kruskal-Wallis test? Should I perform a multiple comparison test (Bonferroni) to confirm the association obtained frmm the one-way ANOVA test? Should I perform additional tests to the ANOVA one?

kjetil b halvorsen
  • 77,844
Alberto Ignacio Herrando
  • 1
  • I would advise posting on Statalist with your data as advice about what to do next depends on what they are like. Bartlett's test isn't taken very seriously now, as even if heteroscedasticity is present there are more direct approaches possible. – Nick Cox Sep 04 '23 at 17:37
  • 1
    Cross-posted as advised at https://www.statalist.org/forums/forum/general-stata-discussion/general/1726020-anova-without-homoscedasticity where there is now much more detail and several answers. – Nick Cox Sep 05 '23 at 11:43
  • And Technically, you should be testing whether or not your residuals are normally distributed, even though what you've done is equivalent... – StatsStudent Sep 06 '23 at 01:04
  • 1
    I don't think statlist is going to offer much better than CV. The OP's question is a question about technique and methods, not programming within Stata. The OP's question could have been equivalently asked without even the mention of Stata. – StatsStudent Sep 06 '23 at 01:08
  • Statalist (not statlist) has already given statistical as well as Stata suggestions and that's what it (we) do. I agree that the Stata mention is immaterial on CV. – Nick Cox Sep 06 '23 at 09:04
  • I was one of those voting for migration to CV. – Nick Cox Sep 06 '23 at 09:38

0 Answers0