# Kruskal wallis

• Jan 16th 2013, 06:16 AM
emme1
Kruskal wallis
Hey,

typ 1 (r)
100 (13)
96 (9)
92 (6)
96 (9)
92 (6,5)

typ2 (r)
76 (2)
80 (3,5)
75 (1)
84 (5)
80 (3,5)

typ3 (r)
108 (15)
100 (13)
96 (9)
98 (11)
100 (13)

My question, is how you do when its 3 of the same. I would do like above.
• Jan 16th 2013, 10:47 AM
ILikeSerena
Re: Kruskal wallis
Hi emme!

Can you elaborate?

You have a series of scores in 3 groups that have already been ranked.
The lowest number has rank 1, and the highest number has rank 15.
All that is left, is to fill in the numbers in the formula for the test statistic.
• Jan 16th 2013, 10:58 AM
emme1
Re: Kruskal wallis
Hey,
Sorry... yes I have 3 groups and I want to rank them, I did the rank but Im not sure its right because I dont know how you rank them when you have the score 100, 3 times.
• Jan 16th 2013, 11:05 AM
ILikeSerena
Re: Kruskal wallis
Ah yes, score 100 would get ranks 12, 13, and 14.
Since it is a tie, each gets the average of these ranks, which is 13.
And that's what you did! Good!
• Jan 16th 2013, 11:20 AM
emme1
Re: Kruskal wallis
Are you sure? Haha.
So I did the score right? And 80, 2 of them gets 3,5 and 96 gets 9?

Thank you for the help!!
• Jan 16th 2013, 11:33 AM
ILikeSerena
Re: Kruskal wallis
Yes! :D

Erm... however, I see you assigned 92, which occurs 2 times, 6 respectively 6,5.
That's not right!
• Jan 16th 2013, 11:43 AM
emme1
Re: Kruskal wallis
Yeah, you're right, I missed that one here... so 92 is 6,5 on both. Except for that, its right?
So if you do a Kruskal test it will be:
k= 12/15(15+1) (44^2/5+15^2/5+61^2)-3(15+1) = 10,82 ?

If thats right, that will make my day! :)
• Jan 16th 2013, 12:02 PM
ILikeSerena
Re: Kruskal wallis
You get my stamp of approval! (Rock)
• Jan 16th 2013, 12:09 PM
emme1
Re: Kruskal wallis
..and then you compare it to a chi critical and reject the Ho: u1=u2=u3.
Haha, I had that question on my exam today and my friends didn't do it like I did, so I thought I was wrong and wanted to check it with you smart guys :)
Thanks a lot!
• Jan 16th 2013, 12:24 PM
ILikeSerena
Re: Kruskal wallis
Erm... my thanks count is still a bit low.
Would you mind increasing it a bit?
• Jan 16th 2013, 12:28 PM
emme1
Re: Kruskal wallis
Haha, I'll do it when I get the result and I can be 100% sure :)
Have a good day!
• Jan 16th 2013, 12:38 PM
ILikeSerena
Re: Kruskal wallis
Ah, but even when you get the results, you can only be about 95% sure.
That's just how statistics works! (Giggle)
CU!
• Jan 22nd 2013, 11:16 AM
ILikeSerena
Re: Kruskal wallis
Well...?
Did you get the result yet?
Are you 100% sure now?
• Jan 23rd 2013, 04:28 PM
martinthonsom
Re: Kruskal wallis
Quote:

Originally Posted by ILikeSerena
Ah, but even when you get the results, you can only be about 95% sure.
That's just how statistics works! (Giggle)
CU!

not really you can adjust to 99%, which software you use for statical analysis, I use statgraphics with 99%

Sydney App Fest