1. ## confused..Kruskal-Wallis!,mutipe Mann-Whitney!,ANOVA!

Hi everyone

I’m doing a research on three different diseases
I’m measuring a protein in theses diseases using an ordinal score
The scores are as follow:
0=absent, 1=low, 2=medium, and 3=high

sample size:
disease(1)=33, disease(2)=26, disease(3)=21

I want to know: Is there any difference in the protein level between the three diseases?

Since my data are ordinal, the best test is Kruskal-Wallis Test
And here is what I got:
Disease (1)>> mean rank=26.06
Disease (2)>> mean rank=48.27
Disease (3)>> mean rank=51.00
Chi-Square=32.336
Df=2
Asymp. Sig=.000

There is an overall difference and it is statistically significant!

Now, how do I know which disease specifically is different from the other? As you know there is no post hoc test or follow up test like in ANOVA! What shall I do?

1-You just need to inspect the mean rank for the three groups and see which one is the lowest and the highest
2-You need to run three Mann-Whitney tests comparing each pair of diseases
3-Do one way ANOVA with a post hoc test considering the data as intervals!!

So where to go? Any suggestion! I know this is may be a controversial issue!
Hope to find the answer supported with a good reference that I can rely on!

PS
My statistics knowledg is really limited i found other too complicated ways which i did not underastand like doing certain manulal calculations
please show me how to do it using SPSS that is the softwear I'm familiar with

2. Xm... Take a look at

Using SPSS for Ordinally Scaled Data:
Mann-Whitney U, Sign Test, and Wilcoxon Tests [link]

3. Originally Posted by mp83
Xm... Take a look at

Using SPSS for Ordinally Scaled Data:
Mann-Whitney U, Sign Test, and Wilcoxon Tests [link]
Thanks
u did not get my point

4. OK. Sorry...

So you want to test the 3 diseases. You can check the pp 5-7 for an illustation on how to do this. Don't worry it's nothing complicated, it goes step by step.

Learning and Understanding the Kruskal-Wallis One-Way Analysis-of-Variance-by-Ranks Test for Differences Among Three or More Independent Groups [link]

If something is confusing you can ask again here. Sorry I;m not spelling everything out myself but I'm pretty occupied at the moment...

5. Dear mp83,
thanks i really appreciate it
Still u did not get my point
I have no problem with the Kruskal-Wallis
as you can see i have done it:
Since my data are ordinal, the best test is Kruskal-Wallis Test
And here is what I got:
Disease (1)>> mean rank=26.06
Disease (2)>> mean rank=48.27
Disease (3)>> mean rank=51.00
Chi-Square=32.336
Df=2
Asymp. Sig=.000

There is an overall difference and it is statistically significant!
MY PROBLEM IS I WANT TO DO A POST-HOC TEST

can sombody show me how to do it using SPSS
OR
do u advise me to considr my data as interval or discrete scale
and just use the parametric ANOVA follwed by its post-hoc

can somebody help me!

6. What the paper describes is what to do after a significant statistic is obtained. That's what you ask for! The procedure only use ranks.

Comparison After the H Statistic
When the obtained value of the H statistic is statistically significant, it indicates that at least one of the groups is different from the others. It does not indicate, however, which groups are different or whether the difference is meaningful, nor does it specify how many of the groups are different from each other. This next procedure, called "multiple comparisons between treatments," constructs pair-wise multiple comparisons to locate the source of significance.~ This procedure tests the null hypothesis that some groups u and v are the same against the alternate hypothesis that some groups u and v are different (Tab. 5). When the sample size is large, these differences are approximately normally distributed. Because there are a large number of differences and because the differences are not independent, however, the comparison procedure must be adjusted appropriately. An inequality is used, and the hypothesis of nodifference among the three groups is tested at the alpha level of significance of .05. The null hypothesis is rejected if the calculated difference among groups is greater than the critical difference. In this example, the difference between treatment groups 1 and 3 (ie, 9.33) is the only difference that is greater than the critical difference (ie, 7.38). It can therefore be concluded that treatment 3 led to a different, and in this case better, result because it provided a greater increase in the range of knee flexion than did treatment 1. (...)

And you should follow the procedure in table 5.You can even do it in a paper. You don't need SPSS.

7. Thanks alot
I see..
but what is the name of this test so i can write it in my paper?
i've red that Dunn's test is widly used and it is designed to maintain the p value at the set level.
do u have any idea how to do it?

8. Dunn's test is what nowadays most peoople call "Bonferroni correction" and I think that it's available to SPSS Base.

You can use the procedure above and cite the paper.

Dunn's Test (The Bonferroni t ). Since the familywise error rate is always less than or equal to the error rate per experiment, , an inequality known as the ...
[http://core.ecu.edu/psyc/wuenschk/docs30/multcomp.doc]

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts