kruskall-wallis analysis problem

mcp

Hello, I'm here to ask for your kind help.

I would like to ask you a couple of questions:

1) I need to test the differences between two independent groups (young and adult) for x-rays abnormalities of the thoracolumbar tract (14 spaces each). Data are ordinal, so I supposed to run the Kruskal-Wallis test to test if a difference occurs in the two group. I didn't choose the two-way independent ANOVA because that test looks for the effect of the parameters on the variable. Is it correct??

2) I tried to run the Kruskal test but once added the factors the bar display an error: residual df = 0 (in attach you will find a screenshot). How can I manage this error? Do you have any suggestion to how to organise the row data (in attach you will find a screen of mine dataset)?

Thank you very much for any help, hope to hear from you soon.

MCP

mcp

datasets+error.pdf

mcp

?this morning I have the doubt I need to run a Mann Whitney... what do you think ?

EJ

Hi mcp,

Sorry for the tardy reply. In JASP, when you select "independent samples t-test" you will be offered the option of doing the Mann-Whitney, so indeed, that's the one you need. This may also explain the error you got when you were trying to run Kruskal (but I'll forward this to the team anyway; perhaps the error message can be made more informative).

Cheers,

E.J.

JohnnyB

Hi mcp,

I think that error is not just concerning the Kruskal Wallis test, but the overall ANOVA. It seems that you only have a single observation per cell of the interaction effect, so when you include that effect, there is no way of calculating the variance per cell. The error message is not very informative, and I think it will be changed in the new release of JASP.

As for your tests: Running the Kruskal Wallis only holds in one-way setups, so with only a single factor. In order to compare levels within a factor to each other (i.e., post hoc tests), you can check the boxes for the appropriate post hoc tests. In your case, I would recommend the Dun post hoc test - from the help file:

• Dunn: This is a non-parametric test that can be used for testing small subsets of pairs. This post hoc test is a follow up for the Kruskal-Wallis test. The p-values are corrected with the Bonferroni and Holm methods

I hope this helps, please let me know if you have additional questions or if anything is unclear =)

Johnny