Richard,
Does the base category matter? Since the coefficients are difficult to
interpret (i.e. the direction of the effect of var_x on outcome 1 is not
determined solely estimated coefficient of var_x for outcome 1 (Wooldridge,
2002 p. 497-498)) wouldn't you usually compute marginal effects, which do
not depend on the base category?
For example, the sign on _mpg_ changes sign when the base category changes
from 3 to 2, but the marginal effect is the same.
. mlogit rep price mpg, base(3) nolog
----------------------------------------------------------------------------
rep78 | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+--------------------------------------------------------------
1 |
price | -.0006468 .000849 -0.76 0.446 -.0023108
.0010173
mpg | -.0197515 .1959901 -0.10 0.920 -.4038851
.364382
_cons | 1.013502 6.695462 0.15 0.880 -12.10936
14.13637
-------------+--------------------------------------------------------------
<snip>
(Outcome rep78==3 is the comparison group)
. mfx , predict(outcome(1)) var(mpg)
Marginal effects after mlogit
y = Pr(rep78==1) (predict, outcome(1))
= .01570292
----------------------------------------------------------------------------
variable | dy/dx Std. Err. z P>|z| [ 95% C.I. ] X
---------+------------------------------------------------------------------
mpg | -.0012526 .00287 -0.44 0.662 -.006868 .004363
21.2899
----------------------------------------------------------------------------
. mlogit rep price mpg, base(2) nolog
----------------------------------------------------------------------------
rep78 | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+-------------------------------------------------------------
1 |
price | -.0005451 .0008602 -0.63 0.526 -.0022309
.0011408
mpg | .0330037 .2163959 0.15 0.879 -.3911245
.4571319
_cons | .6890743 7.076853 0.10 0.922 -13.1813
14.55945
-------------+--------------------------------------------------------------
<snip>
(Outcome rep78==2 is the comparison group)
. mfx , predict(outcome(1)) var(mpg)
Marginal effects after mlogit
y = Pr(rep78==1) (predict, outcome(1))
= .01570292
----------------------------------------------------------------------------
variable | dy/dx Std. Err. z P>|z| [ 95% C.I. ] X
---------+------------------------------------------------------------------
mpg | -.0012526 .00287 -0.44 0.662 -.006868 .004363
21.2899
----------------------------------------------------------------------------
Scott
Wooldridge, Jeffery. "Econometric Analysis of Cross Section and Panel
Data." Cambridge, MA: MIT Press, 2002.
> -----Original Message-----
> From: [email protected] [mailto:owner-
> [email protected]] On Behalf Of Richard. Williams
> Sent: Friday, April 15, 2005 11:32 AM
> To: [email protected]; [email protected]
> Subject: Re: st: RE: mlogit question.
>
> At 09:59 AM 4/15/2005, [email protected] wrote:
> >yes,
> >you should use the option basecategory(#) after the command
> >mlogit depvar [indepvars] [weight] [if exp] [in range] , basecategory(#)
> >anna
>
> I've always found mlogit's practice of using the most frequent category as
> the base category a little annoying. It is easy enough to override, but
> it
> would be nice to be able to say something like basecategory(last) for when
> you want to use the last category as the base but don't happen to remember
> what its value is. It would also be nice if you could permanently set it
> to go with the first or the last value, as I'm always forgetting to
> override the default behavior.
>
>
> -------------------------------------------
> Richard Williams, Associate Professor
> OFFICE: (574)631-6668, (574)631-6463
> FAX: (574)288-4373
> HOME: (574)289-5227
> EMAIL: [email protected]
> WWW (personal): http://www.nd.edu/~rwilliam
> WWW (department): http://www.nd.edu/~soc
>
> *
> * 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/
