2

I'm running ANOVA for 4 clusters to understand is there any difference between these clusters in terms of their reasons for travel. After I checked Levene test, one of my variables was lower than 0.05, but that variable has F(3,324) = 6.35, p=0.000. my question is, can I rely on this ANOVA analysis or not? because Levene test was rejected for this particular variable.

Nick Cox
  • 48,377
  • 8
  • 110
  • 156
Mina
  • 25
  • 1
  • 4
  • Relevant questions on this site: http://stats.stackexchange.com/questions/2492/is-normality-testing-essentially-useless and http://stats.stackexchange.com/questions/61715/choosing-a-statistical-test-based-on-the-outcome-of-another-e-g-normality – Gala Jun 25 '13 at 12:03

1 Answers1

3

Welcome to the site

With a large sample (and yours is fairly large, in these circumstances) any tests can reject H_0 for fairly trivial violations. I suggest looking at parallel box plots of the variable on the four groups. This doesn't give a p value, but it does provide a lot of information. If the variances are very different, it will show up in the box plots.

If you suspect that the rejection of homoscedasticity is correct, then you could try a robust statistic or nonparametric statistic and compare the results of that to the ANOVA to see if it matters. Kruskal Wallis might be a good test for this.

Peter Flom
  • 94,055
  • 35
  • 143
  • 276
  • I tried Kruskal Wallis, and it showed that χ2(3) = 19.306, p = 0.000. so, now that this test confirmed there is a difference, can I use ANOVA in my analysis? I mean scince I used ANOVA for other variables, here where levene test is not confirmed, however can I use ANOVA's result? and about parallel box plots, I created one, but actually I can't interpret it. – Mina Jun 25 '13 at 12:02
  • @Mina You might want to check http://stats.stackexchange.com/questions/48048/how-to-interpret-a-box-plot You can also post your boxplots here (by editing this question or asking a new one). – Gala Jun 25 '13 at 15:06