Hello,
I am trying to use mlogit to perform multinomial logistic regression .
The dependent variable has 6 classes (1 to 6) and the sixth is the base
category. The independent variable is X.
I need to define constraints for the intercepts of the 5 equations (for
categories 1-5), and I therefore use the constraint command.
Here are the constraints that I define :
constraint define 1 [3]_cons = 0.69314718 + 0.5*[1]_cons
constraint define 2 [4]_cons = 0.69314718 + 0.5*[2]_cons
constraint define 3 [5]_cons = 0.69314718 + 0.5*[1]_cons +
0.5*[2]_cons
And below is what STATA returns (I only show the _cons parameters since
they are the ones causing trouble, i.e. the regression parameters for X
are not shown).
Log likelihood = -6290.9233 Pseudo R2 =
0.6955
( 1) - .5 [1]_cons + [3]_cons = .6931472
( 2) - .5 [2]_cons + [4]_cons = .6931472
( 3) - .5 [1]_cons - .5 [2]_cons + [5]_cons = .6931472
1 _cons -2.813621
2 _cons 1.427327
3 _cons -.7136634
4 _cons 1.406811
5 _cons -2.513441
Apparently STATA understands the constraint commands ; however, while
the _cons parameters verify the constraint for the lines 3 and 4 (i.e.,
constraints 1 and 2), the _cons parameter is very different from the one
that should be obtained for the 5th category (given constraint 3, it
should be something like 0.69314718 + 0.5*-2.813621 + 0.5*1.427327 =
.00000018 ).
Does anyone know why the third constraint is not taken into account by
the program ? How can I fix this problem ?
Thank you.
Alexis
*
* 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/