Dear Maarten and Richard,
I attach below the results got when constant is not constrained to be invariant
across alternatives. This time it works even for the predicted probabilities
which are very close to the observed ones. these constants are supposed to
capture fixed effects. You know how to capture random effect as well using the
same mlogit?
thank you a lot, you were really helpful
Edlira
cons drop _all
. cons 1 [2=3]: time cost1
. cons 2 [3=4]: time cost1
. cons 3 [4=5]: time cost1
. mlogit av20 time cost1 if (location>1|location<6|location>6)&purpose==1,
cons(1-3) basecat(1)
Iteration 0: log likelihood = -238.27473
Iteration 1: log likelihood = -233.96107
Iteration 2: log likelihood = -233.76573
Iteration 3: log likelihood = -233.76421
Iteration 4: log likelihood = -233.76421
Multinomial logistic regression Number of obs = 324
LR chi2(2) = 9.02
Prob > chi2 = 0.0110
Log likelihood = -233.76421 Pseudo R2 = 0.0189
( 1) [2]time - [3]time = 0
( 2) [2]cost1 - [3]cost1 = 0
( 3) [3]time - [4]time = 0
( 4) [3]cost1 - [4]cost1 = 0
( 5) [4]time - [5]time = 0
( 6) [4]cost1 - [5]cost1 = 0
------------------------------------------------------------------------------
av20 | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
2 |
time | .012926 .0085442 1.51 0.130 -.0038203 .0296724
cost1 | -.0028968 .0015177 -1.91 0.056 -.0058713 .0000778
_cons | -2.044154 .6947414 -2.94 0.003 -3.405822 -.6824861
-------------+----------------------------------------------------------------
3 |
time | .012926 .0085442 1.51 0.130 -.0038203 .0296724
cost1 | -.0028968 .0015177 -1.91 0.056 -.0058713 .0000778
_cons | -2.848527 .7177557 -3.97 0.000 -4.255302 -1.441752
-------------+----------------------------------------------------------------
4 |
time | .012926 .0085442 1.51 0.130 -.0038203 .0296724
cost1 | -.0028968 .0015177 -1.91 0.056 -.0058713 .0000778
_cons | -4.295446 .8404462 -5.11 0.000 -5.94269 -2.648202
-------------+----------------------------------------------------------------
5 |
time | .012926 .0085442 1.51 0.130 -.0038203 .0296724
cost1 | -.0028968 .0015177 -1.91 0.056 -.0058713 .0000778
_cons | -3.602299 .762463 -4.72 0.000 -5.096699 -2.107899
------------------------------------------------------------------------------
(av20==1 is the base outcome)
. predict p1 p2 p3 p4 p5 if e(sample)
(option pr assumed; predicted probabilities)
(69 missing values generated)
Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
p1 | 324 .7932099 .0648332 .6414537 .9545749
p2 | 324 .1172839 .0367711 .0257635 .2033546
p3 | 324 .0524691 .0164502 .0115258 .0909744
p4 | 324 .0123457 .0038706 .0027119 .0214057
p5 | 324 .0246914 .0077413 .0054239 .0428115
observed probabilities
tab av20 if (location>1|location<6|location>6)&purpose==1
av20 | Freq. Percent Cum.
------------+-----------------------------------
1 | 262 78.92 78.92
2 | 41 12.35 91.27
3 | 17 5.12 96.39
4 | 4 1.20 97.59
5 | 8 2.41 100.00
------------+-----------------------------------
Total | 332 100.00
.
Scrive Maarten buis <[email protected]>:
> At Mon 12/12/2005 10:30 AM, Richard Williams wrote:
> > Also, one thing I find peculiar here is that the equality
> > constraint on the constants isn't working, i.e. the constant in
> > the first 3 equations is 0 while in the last it is non-zero. I
> > would have expected a non-zero and identical value for the
> > constant in all 4 equations.
>
> I agree with Richard that this is odd behaviour, I replicated it
> with the auto dataset. Also the results are close but not quite
> the same as the results from a logistic regression. Oh well, the
> main message from this should be that if you want to estimate a
> logistic regresion, you should use -logistic- and not -mlogit- with lots of
> constraints.
>
> *---------example---------------
> set more off
> sysuse auto, clear
> recode rep78 1=2
> constraint 1 [2=4]: price foreign _cons
> constraint 2 [4=5]: price foreign _cons
> mlogit rep78 price foreign, con(1 2)
> gen dum = rep78 == 3
> logit dum price foreign
> *-------end example-------------
>
> Cheers,
> Maarten
>
> -----------------------------------------
> Maarten L. Buis
> Department of Social Research Methodology
> Vrije Universiteit Amsterdam
> Boelelaan 1081
> 1081 HV Amsterdam
> The Netherlands
>
> visiting adress:
> Buitenveldertselaan 3 (Metropolitan), room Z214
>
> +31 20 5986715
>
> http://home.fsw.vu.nl/m.buis/
> -----------------------------------------
>
>
>
>
>
>
> ___________________________________________________________
> Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with
> voicemail http://uk.messenger.yahoo.com
> *
> * 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/
>
----------------------------------------------------------------
This message was sent using IMP - Facolta' di Economia - Torino.
*
* 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/
