Dear Stata Users,

I have individual defendants data nested within courts and four possible outcomes of the trial. If I split outcome variable in series of dummies I might run the following model e.g. for "fine" as one of four possible outcomes:

melogit fine male age || courtName: , nolog

this is equivalent to:

gsem (fine <- male age M1[courtName]@1) , mlogit

I understand it is irrelevant whether I constrain this term "M1[courtName]" or not, as Stata by default constrain this to unity. My first question is: Why Stata treat simple id variable as a latent one (and whats more as continuous one)? Is this a correct way to indicate nested data?

Next, if I want fit multinomial logit I would go as follows:

mlogit fine male age || courtName: , nolog // However, it is not allowed to fit multilevel model using 'mlogit' (as far as I know).

Hence (and this is my second question), what would be equivalent to the above model in 'gsem'?

this one: gsem (i.outcome <- male age M1[courtName] ) , mlogit

or this one: gsem (i.outcome <- male age M1[courtName]@1) , mlogit

?

Again Stata treats courtName as continuous latent variable, which I don't want. Is it another way to do this?


. gsem (i.outcome <- male age M1[courtName] ) , mlogit
------------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------------+----------------------------------------------------------------
1.outcome | (base outcome)
-------------------+----------------------------------------------------------------
2.outcome |
male | .1250307 .1543864 0.81 0.418 -.1775611 .4276226
age | .0429915 .0066398 6.47 0.000 .0299776 .0560053
|
M1[courtName] | 1 (constrained)
|
_cons | -.4005779 .313917 -1.28 0.202 -1.015844 .214688
-------------------+----------------------------------------------------------------
3.outcome |
male | .4524784 .1867836 2.42 0.015 .0863893 .8185676
age | .0085883 .0081239 1.06 0.290 -.0073342 .0245108
|
M1[courtName] | .3650166 .1427674 2.56 0.011 .0851977 .6448356
|
_cons | -.505497 .2930937 -1.72 0.085 -1.07995 .0689562
-------------------+----------------------------------------------------------------
4.outcome |
male | -1.291661 .1676671 -7.70 0.000 -1.620282 -.9630392
age | .0444199 .0068133 6.52 0.000 .031066 .0577738
|
M1[courtName] | -.8234215 .2766382 -2.98 0.003 -1.365622 -.2812205
|
_cons | -.3104242 .294192 -1.06 0.291 -.8870298 .2661814
-------------------+----------------------------------------------------------------
var(M1[courtName])| .604349 .2992112 .2290133 1.594832
------------------------------------------------------------------------------------

. gsem (i.outcome <- male age M1[courtName]@1) , mlogit
------------------------------------------------------------------------------------
| Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------------+----------------------------------------------------------------
1.outcome | (base outcome)
-------------------+----------------------------------------------------------------
2.outcome |
male | -.1060475 .1562931 -0.68 0.497 -.4123764 .2002814
age | .0393629 .0069448 5.67 0.000 .0257514 .0529744
|
M1[courtName] | 1 (constrained)
|
_cons | .2826794 .3198872 0.88 0.377 -.344288 .9096468
-------------------+----------------------------------------------------------------
3.outcome |
male | .3440249 .1903876 1.81 0.071 -.029128 .7171778
age | .0072179 .0084539 0.85 0.393 -.0093514 .0237872
|
M1[courtName] | 1 (constrained)
|
_cons | -.0462166 .3600873 -0.13 0.898 -.7519747 .6595415
-------------------+----------------------------------------------------------------
4.outcome |
male | -1.184414 .1675455 -7.07 0.000 -1.512797 -.8560307
age | .0469526 .0071332 6.58 0.000 .0329719 .0609333
|
M1[courtName] | 1 (constrained)
|
_cons | .1697173 .3247067 0.52 0.601 -.4666962 .8061308
-------------------+----------------------------------------------------------------
var(M1[courtName])| .5576233 .2498166 .2317398 1.34178
------------------------------------------------------------------------------------