Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.


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

AW: st: xtlogit, margins


From   "Seliger Florian" <seliger@kof.ethz.ch>
To   "'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu>
Subject   AW: st: xtlogit, margins
Date   Mon, 5 Nov 2012 11:53:48 +0000

Dear Maarten, dear Statalist,
You often suggest using odds ratios and - now that I read through the threads -  I find that idea convincing.
However, I am still confused because standard econometric courses don't talk about odds ratios and I haven't heard about them before nor have seen them in any papers. Comments on this issue are welcome.



Just two questions regarding factor variables and margins (just of interest to understand the command):

1.

Suppose I estimate the following models where development and agreements are dummy variables

***********
 logit innopd development#agreements $a 

----------------------------------------------------------------------------------------
                       |               Robust
                innopd |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-----------------------+----------------------------------------------------------------
development#agreements |
                  0 1  |   1.352231   .4246867     3.18   0.001     .5198607    2.184602
                  1 0  |   2.551269   .2361287    10.80   0.000     2.088465    3.014073
                  1 1  |   1.517519   .4206199     3.61   0.000     .6931194    2.341919
(remaining output deleted)



 logit innopd development##agreements $a 

----------------------------------------------------------------------------------------
                       |               Robust
                innopd |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-----------------------+----------------------------------------------------------------
         1.development |   2.551269   .2361287    10.80   0.000     2.088465    3.014073
          1.agreements |   1.352231   .4246867     3.18   0.001     .5198607    2.184602
                       |
development#agreements |
                  1 1  |  -2.385981   .6077377    -3.93   0.000    -3.577125   -1.194837



Why do I get different coefficients for "1 1" ???


2.

If I type:

margins development##agreements

or

margins development#agreements


Is it correct that I will get marginal means?
What are marginal means (formula?), how can I interpret them?







Best, 
Florian







 



 





-----Ursprüngliche Nachricht-----
Von: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] Im Auftrag von Maarten Buis
Gesendet: Sonntag, 4. November 2012 10:12
An: statalist@hsphsun2.harvard.edu
Betreff: Re: st: xtlogit, margins

On Fri, Nov 2, 2012 at 6:15 PM, Seliger  Florian <seliger@kof.ethz.ch> wrote:
> my question is basically whether I can use margins, dydx(*) after 
> xtlogit, re and xtlogit, fe in order to calculate average marginal 
> effects, what margins, dydx(*) will tell me and whether there might be problems in the panel context (the mfx command understates marginal effects in this context, see Cameron, Trivedi, 2009, Microeconometrics using Stata, p.613). I have to say that I am always struggling with the margins command and I am still unsure what all the options do.

With marginal effects one can make a distinction between marginal effects evaluated at average or other meaningful values of the explanatory variables and average marginal effects. The problem with panel data methods is that you don't add all your variables to your model. These unobserved group level variables are captured by the group level variance term. -margins, dydx(*)- now gives you a hybrid version; it fixes the unobserved group level variables at their means, while averaging over the observed explanatory variables. So I don't think that that is quite what you want.

> In addition, I have an interaction effect (xtlogit depvar ind1 ind2 
> ind1Xind2 controls), not using the factor variable language in order 
> to use inteff.

You have to choose: either you use -margins- but than you have to use factor variables for your interactions or you don't use factor variables but than you should not use -margins-

> I am not sure if inteff is outdated (meaning that margins can do the same, but I don't know how) and whether it is appropriate in this case.

For interaction effects I don't think you should use marginal effects at all, see:
<http://www.stata.com/statalist/archive/2010-08/msg00968.html>

Hope this helps,
Maarten

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
Germany

http://www.maartenbuis.nl
---------------------------------

*
*   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/

*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index