There is a method of doing multiple comparisons aftter a KW test. I don't
have the original source, but there is a website from P. Cabillo and J.
Masaro that describes the procedure. See
http://ace.acadiau.ca/math/cabilio/StatLabs/BSPT(02).pdf
p. 27
Al Feiveson
Original Message
From: [email protected]
[mailto:[email protected]]On Behalf Of
[email protected]
Sent: Friday, March 05, 2004 12:20 PM
To: [email protected]
Subject: st: ANOVA Multiple Comparison Questions
Hi! I have a question about ANOVA.
I'd like to see the differences of per capita income(index) for four
different areas. I chose oneway ANOVA and multiple comparison command.
But the problem is the violation of equal variances (please refer to
Bartlett's
test for equal variances below).
My questions are:
1) Can I say that my analysis is valid under the violation of equal
variances?
2) If the violation of equal variances is a problem, is there any way to
get over it?
3) As an alternative, I tried to nonparametic test using KruskalWallis
test.
But "Kwallis" command does not produce multiple comparison for each
group.
Is there any nonparametric ANOVA techniques with multiple comparison?
4) I think this data is unbalanced in terms of # of observation. Is this
problem?
Any comments would be helpful to me.
Thanks.
sugie

.oneway pci_index group [aweight=population], tabulate scheffe
 Summary of pcir0
group  Mean Std. Dev. Freq. Obs.
+
A  1.2176654 .73457797 127574 47
B  .76321118 .43276015 2778506 845
C  1.0260904 .48150203 2989897 767
D  1.0885976 .37953544 6232038 1160
+
Total  1.0000001 .44391706 12128015 2819
Analysis of Variance
Source SS df MS F Prob > F

Between groups 49.4593539 3 16.4864513 91.74 0.0000
Within groups 505.862359 2815 .179702437

Total 555.321713 2818 .197062354
Bartlett's test for equal variances: chi2(3) = 41.7568 Prob>chi2 = 0.000
Comparison of pcir0 by class
(Scheffe)
Row Mean
Col Mean  A B C
+
B  .454454
 0.000

C  .191575 .262879
 0.029 0.000

D  .129068 .325386 .062507
 0.242 0.000 0.018
.kwallis pci_index, by(group)
Test: Equality of populations (KruskalWallis test)
++
 group  Obs  Rank Sum 
++
 A  47  64350.00 
 B  845  801156.00 
 C  767  1.14e+06 
 D  1160  1.97e+06 
++
chisquared = 423.506 with 3 d.f.
probability = 0.0001
chisquared with ties = 423.506 with 3 d.f.
probability = 0.0001
.
*
* 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/
*
* 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/