 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
RE: st: gologit2
For those of you who have lost track of this thread, Maarten
developed some simulations to see how well the brant test does when
ologit's proportional odds assumption is NOT violated. Ideally, the
null should be rejected 5% of the time. As his simulations using the
ologit and brant commands showed, the more vars in the model, the
more likely the brant test was to incorrectly reject the null:
# of Xs | % reject H0
----------------------
1 | 5.18
2 | 5.55
3 | 5.19
4 | 5.49
5 | 5.61
6 | 5.39
7 | 5.73
8 | 5.88
9 | 6.09
10 | 6.33
12 | 7.06
14 | 7.58
16 | 8.67
18 | 9.75
20 | 11.85
----------------------
The -omodel- command is an alternative to brant. Here are the
simulation results using -omodel- :
# of Xs | % reject H0
----------------------
1 | 5.19
2 | 5.14
3 | 4.97
4 | 4.86
5 | 4.80
6 | 4.64
7 | 4.90
8 | 4.77
9 | 5.01
10 | 5.59
12 | 6.36
14 | 7.06
16 | 8.91
18 | 9.91
20 | 11.71
----------------------
The omodel test appears to be slightly more conservative than the
brant test, but overall the pattern is the same; once you get past 10
vars, the test is more and more likely to reject the null, at least
in this particular simulation.
I'll try it next using the likelihood ratio test that can be
generated using gologit2 and report back later. My guess is it will
do about the same, although I hope the fact that it is a LR test lets
it do a little better.
-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME: (574)289-5227
EMAIL: [email protected]
WWW: http://www.nd.edu/~rwilliam
*
* 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/