Re: st: Help about Ranksum - porder option

[Richard Goldstein <richgold@ix.netcom.com>]

porder is taking all possible pairs (control=yes paired with each
control= no) and asking in what percent of those pairs is No larger than YES; 
this does not appear to be the question you are asking; if I correctly
understand your question, I think you want a parametric procedure rather than 
the rank-sum test

hope this helps
Rich

Óscar Becerra-CERAC wrote:


> Hi all,
>
> I am applying a Wilcoxon test for the equality of distributions, between two
> control groups with the ranksum command. But I have a doubt about the
> interpretation of the result of this test. The obtained result is:
>
> ranksum series , by(control) porder
>
> Two-sample Wilcoxon rank-sum (Mann-Whitney) test
>
>   control    |      obs    rank sum    expected
> -------------+---------------------------------
>           NO |     2284     3280197     3338066
>          YES |      638      990306      932437
> -------------+---------------------------------
>     combined |     2922     4270503     4270503
>
> unadjusted variance   3.549e+08
> adjustment for ties  -5118462.8
>                      ----------
> adjusted variance     3.498e+08
>
> Ho: series(control==NO) = series(control==YES)
>              z =  -3.094
>     Prob > |z| =   0.0020
>
> P{series(control==NO) > series(control==YES)} = 0.460
>
>
> According with the p - value, I understand that the distribution between –
> groups changes, but I can’t understand the result obtained by the p order
> option. What I really want to know is if when my control measure is on it
> has a positive impact in the behaviour of series. Can you help me with this?
>
> Thanks,
>
>
>
> Óscar Becerra
> Researcher
>  
> Conflict Analysis Resource Center
> Carrera 10 N° 65 - 35 Of. 603
> Tels: +(571) 345 85 72/3
> www.cerac.org.co
*
*   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/