Your answer strengthens my previous argument.
-makematrix- just cannot fulfil all the expectations
you have of it. If you omit the spaces in
max(chiprob(r(df), r(chi2_adj) + 1e-20),.0001)
to
max(chiprob(r(df),r(chi2_adj)+1e-20),.0001)
you might get -makematrix- to behave the way you want,
as the argument of -from()- is parsed on spaces,
but you are still ignoring my strong advice not to do it
this way.
I am at a loss to see why you prefer this route to
that of a program designed to tackle your problem
directly. No programming on your part is required.
Nick
n.j.cox@durham.ac.uk
Herve STOLOWY
> Dear Nick:
>
> Sorry for this late reply. I tried your program and it works
> perfectly. I thank you very much.
>
> Although I perfectly read what you wrote about -makematrix-,
> I still like this command and my level of programming is not
> good enough to write the program you designed for me.
>
> So, I would have an additional question, hopefully more simple.
>
> With a ranksum test, I learnt from prior postings how to get a matrix:
>
> makematrix m1, from(r(z) 2*normprob(-abs(r(z)))) label :
> ranksum prov_ass, by( sodas2cl)
>
> I woud like to do the same with -kwallis-. In your program, I
> found the following formula which seems to be the computation
> of the p-value of the kwallis:
>
> max(chiprob(r(df), r(chi2_adj) + 1e-20),.0001)
>
> Then, I wrote the following command with -makematrix-:
>
> makematrix m2, from(r(chi2_adj) max(chiprob(r(df),
> r(chi2_adj) + 1e-20),.0001) ) label : kwallis prov_ass, by(sodas3cl)
>
> Unfortunately, it does not work. I get the following error message:
>
> invalid '.0001'
> r(198);
>
> I obviously did something wrong.
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/