Re: NPAR TESTS

There's a Johnkheere-Terpstra on p. 864 of spssbase.pdf, but I don't know if 
it's exactly the same thing I'm thinking of.  What I'm thinking of is a 
statistic where:

A := 0;
for group := 1 to K-1 do
begin
	for case := 1 to n do
	begin
		for right_group := group+1 to K do
		begin
			for right_case := 1 to n do
			begin
				if observation[right_group, right_case] > observation[group,case] then 
					A := A + 1;
			end;
		end;
	end;
end;

do_some_calculations(A, K, n);

where K is the number of unrelated groups, n is the number of cases in each 
group, which must be the same, and the data are arranged in a 2-d array.

It's intended as an extension to K-W in cases where you expect a specific 
order for group medians, so it's more powerful.  Nevertheless, it's the most 
unusual NPAR TEST I can think of for the moment.

But I don't know if it's exactly the same as the J-T command in the manual.  
If it is, it's described badly.  Maybe J-T in SPSS is a more generalised form 
since it doesn't mention a restriction on n.  I'll look it up sometime next 
week. 

- M.

On Monday 29 March 2004 20:55, Ben Pfaff wrote:
> Michael Kiefte <[email protected]> writes:
> > So, I can't think of any strange tests except for one: the Johnkheere
> > trend test, and I have some ideas for that.
>
> I don't see this test mentioned in the SPSS Syntax Reference
> guide I here (for version 11.5).  Can you tell me more?