Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: -margins- after -gsem- multinomial logit in Stata 13


From   Katia Covarrubias <katia.covarrubias@graduateinstitute.ch>
To   statalist@hsphsun2.harvard.edu
Subject   st: -margins- after -gsem- multinomial logit in Stata 13
Date   Tue, 29 Oct 2013 15:19:50 +0100

Dear Statalist,

As part of my work with a two-year household survey panel dataset, I
am working with the new -gsem- command in Stata 13 for Mac for
generalized structural equation models which allows for a multinomial
logit to be estimated with random effects. The documentation on
obtaining the marginal effects after -gsem- estimation is sparse and I
am having trouble with specifying -margins- as a post-estimation
command.

My general specification is:

----
gsem (2.activity 3.activity 4.activity <- 1.dvar1 1.dvar2 cvar1
M1[hhindidn] ), mlogit
----

in which activity is my polychotomous dependent variable taking values
1, 2, 3, 4; dvar1 and dvar2 are dummy variables coded as 0/1; and
cvar1  s a continuous explanatory variable. I have included an
individual random effect (M1[hhindidn]) and indicate that the model is
a multinomial logit with the option ("mlogit").

According to  the documentation for -gsem-, -margins- is valid as a
post-estimation command. However, the standard approach for specifying
-margins- after a multinomial logit model does not seem to work after
-gsem-.

For example, I have tried running the following commands for
estimating the marginal effects of my key variable dvar1, and receive
the errors indicated after each command:

----
. margins dvar1, predict(outcome(2))
invalid outcome() option;
depvar 2 not found
----

. margins dvar1, predict(outcome(activity)
prediction is a function of possibly stochastic quantities other than e(b)
----

. margins dvar1, predict(outcome(2.activity))
prediction is a function of possibly stochastic quantities other than e(b)
----

. margins dvar1, predict(outcome(activity(2)))
invalid outcome() option;
depvar activity3(2) not found
----

Would anyone have any insights on how to correctly specify -margins-
following a -gsem, mlogit- regression or if there might be some other
alternatives for obtaining the marginal effects?

Many thanks,
Katia

--
Katia Alejandra Covarrubias
Research Assistant, Centre for International Environmental Studies
PhD Candidate, Development Economics
katia.covarrubias@graduateinstitute.ch
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index