I'm looking to compare three independent samples when both normality and homogeneity of variances assumptions are violated, I attempted to use kruskal-wallis test but I read that if homogeneity of variances is violated then it can no longer work, so which test should I use in this case?
Possibly the assumption that your response variable is normal distributed is not very sensible. Do you have a better idea (ideally theoretically founded) how you could model the distribution of the response variable?
@Jochen Wilhelm, you mean something like ordinary linear regression?
No, I mean that you should find out what distributional assumptions makes sense for your response variable.
It might be helpful to sit together with a local statistician to clarify your research design, the scientific question you want to address and what statistical model is most suited to help you (what includes thinking about sensible assumptions about the distribution). This may also help you not to draw wrong conclusions from the numbers being spit out by the statistical tests.
well, that's what I'm trying to achieve by posting my question on researchgate.
Yes, and as I see it - based on year of experience - this is not something you will get solved satisfactorily on RG. Therefore I suggest to sit together with a statistician and really discuss your project. If you catch a good one, you have the chance to learn a lot that will help you in future, too.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA