Re: NPAR TESTS

On Friday 26 March 2004 18:46, you wrote:
> On Fri, Mar 26, 2004 at 11:32:05AM -0400, Michael Kiefte wrote:
>      I'm going to have a go at NPAR TESTS unless someone objects.  I
>      probably won't  work on it this weekend though.
>
> I was going to work on this one myself.  But I haven't started
> anything. So please feel free.
>
> Have you thought about how to calculate the Wilcoxon's signed rank
> statistic?  It's O(n2^n) .....

Sorry, I guess you meant the distribution function.  There is a normal 
approximation for large sample sizes:

z = \frac{T - \frac{n(n + 1)}
                      {4}
         }
         {\sqrt{ \frac{n(n+1)(2n+1)}
                      {24} 
               }
         }

I don't know what "large" means.  There's some variants that do continuity 
corrections and some that correct for large numbers of ties.  It might take a 
bit of optimising in order to determine which method is best for 
fast/accurate results for a given precision.  Sometimes SPSS says p < 0.05 
and sometimes it says p = 0.02.

I'll look around for other algorithms too.

As for division of labour, we should make some decisions. I'm keen to do just 
about anything. I still haven't figured out the SPSS syntax or the calling 
structure in PSPP, so I'm sure to run into a lot of problems that I'll need 
help with.  Keep that in mind if you want to assign anything to me.

Or I can lay low and mark exams...

- M.