## Kruskal-Wallis H adjusted for ties in SPSS?

### Kruskal-Wallis H adjusted for ties in SPSS?

Is there a way in SPSS to automatically calculate Kruskal-Wallis H' test
stastistic adjusted for tied ranks (I know the formula)? It seems that the
ordinary H (or chi square in SPSS terminology) is the only one provided. I
am having 5 x 3 table (one nominal vs one ordinal varible) where I would
like to analyze an association with the this specific test statistic.

any help is greatly appreciated
Roberto Oliveri

### Kruskal-Wallis H adjusted for ties in SPSS?

On Tue, 16 Jul 2002 22:40:38 +0200, "Roberto S. Oliveri, M.D."

> Is there a way in SPSS to automatically calculate Kruskal-Wallis H' test
> stastistic adjusted for tied ranks (I know the formula)? It seems that the
> ordinary H (or chi square in SPSS terminology) is the only one provided. I
> am having 5 x 3 table (one nominal vs one ordinal varible) where I would
> like to analyze an association with the this specific test statistic.

page 499 of the 1990 Reference Guide:

"NPAR TESTS  K-W = varlist BY variable(value1,value2)

"The display shows the number of valid cases, mean rank of the
variable in each group, chi-square, probability, and chi-square and
probability after correcting for ties."

--

http://www.pitt.edu/~wpilib/index.html

### Kruskal-Wallis H adjusted for ties in SPSS?

I am using SPSS 11.0. Strangely enough this version provides only one single
tied ranks.

I am not quite sure then, how to calculate H' from H (chi-square). However,
Roberto Oliveri

> On Tue, 16 Jul 2002 22:40:38 +0200, "Roberto S. Oliveri, M.D."

> > Is there a way in SPSS to automatically calculate Kruskal-Wallis H' test
> > stastistic adjusted for tied ranks (I know the formula)? It seems that
the
> > ordinary H (or chi square in SPSS terminology) is the only one provided.
I
> > am having 5 x 3 table (one nominal vs one ordinal varible) where I would
> > like to analyze an association with the this specific test statistic.

> page 499 of the 1990 Reference Guide:

> "NPAR TESTS  K-W = varlist BY variable(value1,value2)

> "The display shows the number of valid cases, mean rank of the
> variable in each group, chi-square, probability, and chi-square and
> probability after correcting for ties."

> --

> http://www.pitt.edu/~wpilib/index.html

### Kruskal-Wallis H adjusted for ties in SPSS?

Though the manual doesn't mention it explicitely it seems that chi-square
(or H statistic) provided in the output will automatically be the corrected
one if ties are present.

Roberto Oliveri

,Dear Sirs

I use Spss version 7.00 for Windows in order to test whether there is a
linear trend between 3 different groups.
I use the Noneparametric test command with the K independent tests
subcommand.
As a result i receive the Kruskal-Wallis value but not the Cuzick's value
for linear trend.
Does anybody know how to obtain the Cuzick's test value?
Thank you
Israel Schorr