One way ANOVA vs Kruskal Wallis

n.lu

New Member
#1
Hi,

I am comparing four groups (measurements of bone thickness and other bone parameters). The distributions are normal (Kolmogorov–Smirnov test) and variances equal between groups (Levenes test). I have 7 samples in each group.
As I see from the literature this would call for a parametric test. But the results from Kruskal Wallis make more sense than those from ANOVA - I get significant differences on more comparisons whit KW than with ANOVA and biologicly and from what we have observed during analysis those results from KW would make more sense.
So my question is if it is ok to use a nonparametric test in this case or is that wrong?

Best,
Nina
 
Last edited:

Karabiner

TS Contributor
#2
I get significant differences on more comparisons whit KW than with ANOVA
Do you mean, from the respective global tests, or from the pairwise comparisons? If the latter is the case, how were these pairwise comparisons performed?
So my question is if it is ok to use a nonparametric test in this case or is that wrong?
Why did you perform K-W although you knew from literature that ANOVA was o.k. here?

With kind regards

Karabiner