I want to compare variances of several independent samples. Levene's test is significant. How can I perform post hoc comparisons of variances (particularly, in SPSS)? I have two ideas:
1) Run several pairwise Levene’s tests using manual Bonferroni or Sidak correction.
2) Save unstandartized residuals into the dataset and use they as outcome variable in One-Way ANOVA with post hoc tests. If this method is correct, may I use Tukey's or Games-Howell's test (I don’t like conservative Bonferroni correction)?
Do you have better ideas?
I'd use #1. Find the individual pairwise p values. Probably use Holm's procedure rather then those suggested (as you say, Bonferroni is conservative), but depends why you are doing this.
Daniel Wright , I am studying representations of national traditions among Koreans in Uzbekistan. There are three samples: young Koreans, living in Uzbekistan, young Koreans, working in South Korea, old Koreans, living in Uzbekistan. I suppose that representations of old people is more homogeneous, that representation of young people.
I agree with Daniel Wright. If you go with his approach you may find this attachment useful. Best, David Booth
- Badges
- Science method