Statalist


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

Re: RIF: st: mfx with xtlogit


From   Johannes Geyer <JGeyer@diw.de>
To   statalist@hsphsun2.harvard.edu
Subject   Re: RIF: st: mfx with xtlogit
Date   Tue, 26 Feb 2008 12:20:00 +0100

>the mfx default does not correspond to the median of the dependent 
variable evaluated at sample means of >independent variables? to evaluate 
at the mean I guess it is necessary to predict the mean of the dependent 
>variable before mfx

from -mfx- helptext:
Exactly what mfx can calculate  is determined by the previous estimation 
command and the  predict(predict_option) option.  The values at which the 
marginal effects  or elasticities are to be evaluated is determined by the 
at(atlist) option.  By default, mfx calculates the marginal effects or 
elasticities  at the means of the independent variables using the default 
prediction  option associated with the previous estimation command.



Johannes

----------------------
Johannes Geyer
Deutsches Institut für Wirtschaftsforschung (DIW Berlin)
German Institute for Economic Research 
Department of Public Economics
DIW Berlin
Mohrenstraße 58
10117 Berlin
Tel: +49-30-89789-258



"Mussida Chiara" <chiara.mussida@unicatt.it> 
Gesendet von: owner-statalist@hsphsun2.harvard.edu
26/02/2008 12:11
Bitte antworten an
statalist@hsphsun2.harvard.edu


An
<statalist@hsphsun2.harvard.edu>
Kopie

Thema
RIF: st: mfx with xtlogit






the mfx default does not correspond to the median of the dependent 
variable evaluated at sample means of independent variables? to evaluate 
at the mean I guess it is necessary to predict the mean of the dependent 
variable before mfx
 
chiara

                 -----Messaggio originale----- 
                 Da: owner-statalist@hsphsun2.harvard.edu per conto di 
Johannes Geyer 
                 Inviato: mar 26/02/2008 11.44 
                 A: statalist@hsphsun2.harvard.edu 
                 Cc: 
                 Oggetto: Re: st: mfx with xtlogit
 
 

                 Dear Alejandro,
 
                 it happens that coefficients are significant and marginal 
effects not.
                 They simply depend on all other coefficients which makes 
it hard to
                 predict their significance - e.g. it might be important 
to decide whether
                 to include insignificant coefficients of other variables 
in your
                 calculation.
                 You have to decide over which marginal effect to 
calculate - did you try
                 -margeff- by Tamas Bartus? In my view it is more flexible 
than mfx. You
                 can e.g. calculate average marginal effects, i.e. 
calculate the mean
                 effect in your sample (margeff can do this) or simply 
evaluate your
                 function at sample means (mfx default). If your sample is 
small you could
                 also think of bootstrapping standard errors of marginal 
effects.
 
                 Hope that helps,
 
                 Johannes
 
                 ----------------------
                 Johannes Geyer
                 Deutsches Institut für Wirtschaftsforschung (DIW Berlin)
                 German Institute for Economic Research
                 Department of Public Economics
                 DIW Berlin
                 Mohrenstraße 58
                 10117 Berlin
                 Tel: +49-30-89789-258
 
 
 
                 Alejandro Delafuente <alejandro.delafuente@sant.ox.ac.uk>
                 Gesendet von: owner-statalist@hsphsun2.harvard.edu
                 26/02/2008 11:20
                 Bitte antworten an
                 statalist@hsphsun2.harvard.edu
 
 
                 An
                 statalist@hsphsun2.harvard.edu
                 Kopie
 
                 Thema
                 st: mfx with xtlogit
 
 
 
 
 
 
                 Dear statalisters,
 
                 I estimated marginal effects after running xtlogit with 
fixed effects as
                 follows:
 
                 xtlogit depvar independentvars, i(folio) fe
                 mfx, predict(pu0)
 
                 Where depvar indicates reception of transfers. All the 
marginal effects of
 
                 interest are insignificant (most of which are dummy 
variables), except for
                 one
                 variable which has few observations across my 3-round 
panel. This came as
                 a
                 surprise for two reasons:
                 1) the xtlogit coefficients for some of those same
                 variables were significant;
                 2) I had run probit over the pooled sample (ie: dprobit 
depvar
                 indepvars)and
                 xtprobit (ie: xtprobit depvar indepvars, re) BEFORE 
xtlogit  and some of
                 those
                 same coefficients were highly significant as well.
 
                 My understanding is that there isn't one single way for 
capturing marginal
 
                 effects after xtlogit, but am a bit puzzled with what 
I've found thus far
                 and
                 wonder whether other commands for estimating mfx after 
xtlogit exist? Or
                 any
                 advice as to why this insignificancy might be taking 
place?
 
                 Thanks,
 
                 Alejandro
                 --
                 Alejandro de la Fuente
                 Department of International Development/QEH
                 University of Oxford, Mansfield Road, Oxford OX1 3TB
                 Tel: 01865 281836
 
                 *
                 *   For searches and help try:
                 *   http://www.stata.com/support/faqs/res/findit.html
                 *   http://www.stata.com/support/statalist/faq
                 *   http://www.ats.ucla.edu/stat/stata/
 
 
 
                 *
                 *   For searches and help try:
                 *   http://www.stata.com/support/faqs/res/findit.html
                 *   http://www.stata.com/support/statalist/faq
                 *   http://www.ats.ucla.edu/stat/stata/
 


Attachment: winmail.dat
Description: Binary data




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