[Limdep Nlogit List] Ordered probit and marginal effects

Andreas Drichoutis adrihout at aua.gr
Tue Apr 11 19:17:02 EST 2006

Dear all,


I'm estimating an ordered probit and were puzzled by the marginal effects.
The dependent variable takes values from 0 to 4. The marginal effects and
the corresponding t-ratios for the last category look something like this
(I've included the command line):



These are the effects on Prob[Y=04] at means.

 Constant     .0000000000 ........(Fixed Parameter)........

 WORKPHDM  .1874593720E-01  .65198239E-02    2.875   .0040     1.2707692

 INVOLV   -.1441454829E-01  .50133698E-02   -2.875   .0040     3.6123077

 WORKWALK -.8324789601E-02  .82463636E-02   -1.010   .3127     1.5292308

 *PLANNER  .7666590987E-01  .26456434E-01    2.898   .0038     .71076923

 STRAIN    .6918060216E-03  .12568039E-01     .055   .9561     .88615385

 *EXER     .8775940423E-01  .36526920E-01    2.403   .0163     .31076923

 *NOSMOKE -.1098451110E-01  .46540021E-01    -.236   .8134     .41230769

 *SMSTOP   .4750989226E-03  .44843239E-01     .011   .9915     .16923077



The t-ratios are always the same (in absolute values) for the first two
variables. As a matter of fact the t-ratios will be the same for the first
two variables no matter what these variables are. I've tried to replace the
first two variables with other variables and the t-ratios will always be
identical for the first two. If I change the order of the independent
variables the estimated t-ratios change. Is there a bug here??



Andreas Drichoutis



More information about the Limdep mailing list