Re: st: Re: interpretation

[Judith Abrams <abramsj@karmanos.org>]

I believe there is an error in kwallis2. I wrote to the author and he 
disagreed. The reason for my concern is the null case.  According to my 
understanding, if there are only 2 groups, kwallis and kwallis2 should 
produce the same pvalues, but they do not. For example, using the auto data

kwallis trunk, by(foreign)

Test: Equality of populations (Kruskal-Wallis test)

 +---------------------------+
 |  foreign | Obs | Rank Sum |
 |----------+-----+----------|
 | Domestic |  52 |  2218.00 |
 |  Foreign |  22 |   557.00 |
 +---------------------------+

chi-squared =    10.045 with 1 d.f.
probability =     0.0015

chi-squared with ties =    10.134 with 1 d.f.
probability =     0.0015

**
kwallis2 trunk, by(foreign)

One-way analysis of variance by ranks (Kruskal-Wallis Test)

foreign  Obs   RankSum  RankMean
--------------------------------
 0       52   2218.00     42.65
 1       22    557.00     25.32

Chi-squared (uncorrected for ties) =    10.045 with    1 d.f. (p = 0.00153)
Chi-squared (corrected for ties)   =    10.134 with    1 d.f. (p = 0.00146)

Multiple comparisons between groups
-----------------------------------
(Adjusted p-value for significance is 0.025000)

Ho: trunk(foreign==1) = trunk(foreign==2)
   RankMeans difference =     17.34  Critical value =     10.72
   Prob = 0.000764 (S)


--

Judith Abrams
Barbara Ann Karmanos Cancer Institute
Wayne State University


*
*   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/