Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: gllamm level 3 model


From   Alexander Staus <alexander.staus@uni-hohenheim.de>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: gllamm level 3 model
Date   Wed, 24 Sep 2008 13:36:59 +0200

Hi Claudio,

I followed mainly gllamm manual and some examples on their website:
http://www.gllamm.org/examples.html

I think you can also try the "mixlogit" command in stata for your model.

Greetings,
Alex


Claudio Cruz Cazares schrieb:
Hi Alex,

Thank you so much for your answer. I was following gllamm manual examples for multinomial logit, that's why I used "fam (binom)". I try again as you said and I got a satisfactory output. I don't really understand why.

Could you recommend me a specific literature on this subject?,
Thank you again,

Claudio



----- Mensaje original -----
De: Alexander Staus <alexander.staus@uni-hohenheim.de>
Fecha: Martes, Septiembre 23, 2008 2:56 pm
Asunto: Re: st: gllamm level 3 model


Hi Claudio,

I think you are right with your level 2 model, but why do you use "fam(binom)"? Try:

gllamm AID PERSOC IPNCL1 IPNFL1 IPNML1 InFIDT, i(Identi) link (mlogit) base(1) nocorr adap nip(4) nrf(3) eqs (a2 a3 a4) trace

Good luck,
Alex



Hi to all,

I’m running a multinomial regression model with random effects
using gllamm and some doubts have come to me. I really would appreciate your help.

I have a panel data of 17785 observations of 2475 firms during a
14 years period with a polytomous dep_var with four categories. That would be a level 3 model, observation nested in firms nested in years, isn’t it?

When I try the next syntax, the level 1 and level 2 have the
same number of observations!!

gllamm AID PERSOC IPNCL1 IPNFL1 IPNML1 InFIDT, i(Identi year)
link (mlogit) base (1) fam (binom) nocorr adap nip (4) nrf (2 1) eqs (a2 a3 a4) trace

number of level 1 units = 17785
number of level 2 units = 17785
number of level 3 units = 14

Moreover, I’m trying to include random intercepts for each
alternative of the dependent variable at level 2 (firms) using eqs (a2 a3 a4) nrf (3 0) but this option is not allowed, since gllamm need to “split” the random effect in level 2 and 3. That is:

...eqs (a2 a3 a4) nrf (2 1)

Random effects information for 3 level model
-----------------------------------------------------------------
------------

***level 2 (Identi) equation(s):
(2 random effect(s))
standard deviation for random effect 1
Ide1_1 : a2
standard deviation for random effect 2
Ide1_2 : a3

***level 3 (year) equation(s):
(1 random effect(s))

standard deviation for random effect 3
yea2_1 : a4

But I’m not sure how to interpret this and if it is
theoretically correct.

I had try to make a level 2 model by not considering the years
and then the number of units in level 1 and level 2 are correct.

number of level 1 units = 17785
number of level 2 units = 2475

But what about the years???. Could I control for this by
including them in the fixed part of the model?

Thank so much in advance
Claudio
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index