Kruskall Wallis test

[Byrd Sellers, Johnita]

Hello All,

 

I want to compare the mean scores of three different groups.  However, I would need to use the Kruskal-Wallis test because the variables are NOT normally distributed.  I know that the results of the Kruskal-Wallis test gives you the mean of the ranks and not the means of the raw data values.  If I wanted to test if a significant difference exists, I realize that I would need to use a nonparametric test.  But, if I wanted to show the means of the raw data, could I use another method to show this?  In other words, is it ok to state the p-value from one test but use the means obtained from a different test?  Any advice that you can give regarding this issue would be greatly appreciated.  Thanks in advance for all your help…

 

Johnita

 

[Michael Kruger]

Johnita,

One of the options fro the K_W output is to show the descriptives which will give means and SDs. The mean and SD are NOT statistical tests. They are merely descriptive statistics that provide a summary description of a data set. It is perfectly acceptable to cite them while using non-parametric statistical tests for your hypothesis testing.

M. Kruger

__________ Information from ESET NOD32 Antivirus, version of virus signature database 3865 (20090218) __________

The message was checked by ESET NOD32 Antivirus.

http://www.eset.com