st: RE: update - trouble with computing marginal effects after a Logit or an Ordered Logit model - Please HELP

["William Buchanan" <william@williambuchanan.net>]

Hi Jeremy,

Since you're using Stata 11, I would first suggest reading [U]11.4.3 Factor
Variables.  You can simplify your syntax quite a bit just by looking there
and learning how to use factor variables. Then your syntax might look like:

logit mfirstchoice i.mstudy mhomme mchiefwageearner i.mage i.mintpol i.mpol
i.mincome i.s003##i.s002, vce(cluster s003)

One of the issues that I think you're having might be the result of
multicollinearity, since it seems like you're including all of the possible
categories of some factors in the model.  Using factor variable notation can
help you to reduce that happening quite a bit, since Stata will
automatically exclude a reference group.  

Once you fit your model (like the one above), then you can use the -margins-
command which is part of official Stata (see [R] margins for details about
the syntax or type -help margins-).  There are multiple options available
and the command is very flexible.  Also, you should include the results that
you're getting from Stata.  This article from one of the StataJournal
Associate Editors (http://www.maartenbuis.nl/publications/interactions.pdf)
might be helpful to look at.

HTH,
Billy

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Jeremy Franklin
Sent: Thursday, July 26, 2012 1:33 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: update - trouble with computing marginal effects after a Logit
or an Ordered Logit model - Please HELP

Dear all, 

Here is my little trouble regarding assessing marginal effects after running
a logit model on STATA 11: 

First, please not that I am a beginner with STATA and what follows might
sound ridiculous 

For my master degree thesis I decided to test for the role of education
level in assessing the importance of fighting inflation. 

Here is my final regression formula: 

xi: logit mfirstchoice mstudymid mstudyhigh mhomme mchiefwageearner
mage28_37 mage38_47 mage48_57 mage58 mintpollow mintpolmid mintpolhigher
mpolleft mpolright mincomemid mincomehigh i.s003 i.s002 i.s003*i.s002,
vce(cluster s003) 

Note that:
- all these variables are binaries
- i.s003 controls for the country effects
- i.s002 controls for the wave (time) effects
- i.s003*i.s002 controls for the country-wave effects
- vce(cluster s003) characterizes the error term 

I have the results but my thesis coordinator told me that the coefficients
of logit regressions cannot be interpreted like coefficients of a linear
regression. Therefore, he suggested me to check for the marginal effects at
the median. The purpose of this last procedure (if I remember well) is to
see the marginal effects of one individual having a higher level of
education relative to lower or middle education level.

I googled everything, I read the mfx and the margins search page on stata
about hundred times, I tried hundreds of formulas, both with mfx and margins
but i still cannot find the correct one in order to interpret my results. 

Can ANYONE help me please? It is very important

ps: a robustness test included in my thesis uses the following formula (this
time with ologit as far as the dependent variable takes multiple ordered
outcomes): 

xi: ologit minflation mstudymid mstudyhigh mhomme mchiefwageearner mage28_37
mage38_47 mage48_57 mage58 mintpollow mintpolmid mintpolhigher mpolleft
mpolright x047 i.s003 i.s002 i.s003*i.s002, vce(cluster s003)

So if it is easier to check for the marginal effects with a ordered logit
please let me know.

Kind regards;

Franklin Jeremy

*
*   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/

*
*   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/