[Limdep Nlogit List] random parameters probit
William Greene
wgreene at stern.nyu.edu
Thu Apr 2 22:20:49 EST 2009
Andreas:
(1) The "marginal effect" for the constant term should be ignored.
It should not even be reported.
(2) Marginal effects are derivatives, not probabilities. They are
not bounded by 0 and 1 (of course, they can be negative.). If you
x variable is very small, its coefficient will be very large. The
scale factor is the density, which will be about 0.2 regardless.
The ME is beta*scale factor, which can be anything. The interpretation
is that the ME is a change in the probability per one unit change
in the variable. A one unit change in the variable may be very large
and not be a realistic experiment. That is why one occasionally
reports elasticities of probabilities rather than derivatives or
semi elasticities, dP/dlogX which corresponds to the change in the
probability for a one % change in X. You might also consider a one
standard deviation change in X.
/Bill Greene
----- Original Message -----
From: "Andreas Drichoutis" <adrihout at aua.gr>
To: limdep at limdep.itls.usyd.edu.au
Sent: Thursday, April 2, 2009 4:50:29 AM GMT -05:00 Columbia
Subject: [Limdep Nlogit List] random parameters probit
Dear all,
I'm estimating a random parameters probit model for panel data using the
below syntax:
namelist ;
x=period,treat_pr,treat_tr,per_pric,gender,age,hsize,totfee,econpos1,econpos
2,econpos3,econpos4 $
skip $
calc ; ran(12345) $
probit ; lhs=pr12 ; rhs=one,x ; pds=10;rpm ;
fcn=one(n),treat_pr(n),treat_tr(n),per_pric(n);marginal ;HALTON $
The marginal effects are exhibited below. Shouldn't marginal effects be
bounded by 0 and 1? What is the interpretation of the constant with a value
of -1.81?
Best,
Andreas Drichoutis
+-------------------------------------------+
| Partial derivatives of expected val. with |
| respect to the vector of characteristics. |
| They are computed at the means of the Xs. |
| Conditional Mean at Sample Point .1488 |
| Scale Factor for Marginal Effects .2319 |
+-------------------------------------------+
+--------+--------------+----------------+--------+--------+----------+
|Variable| Coefficient | Standard Error |b/St.Er.|P[|Z|>z]| Mean of X|
+--------+--------------+----------------+--------+--------+----------+
|PERIOD | .02031*** .00460608 4.411 .0000 5.5000000|
|GENDER | -.21272*** .05148147 -4.132 .0000 .3943662|
|AGE | .02178* .01251948 1.740 .0818 20.732394|
|HSIZE | -.04473** .02031084 -2.202 .0276 4.3802817|
|TOTFEE | .03920* .02344034 1.672 .0944 16.711268|
|ECONPOS1| .63131*** .15221410 4.148 .0000 .2676056|
|ECONPOS2| .86991*** .17124291 5.080 .0000 .2676056|
|ECONPOS3| .52777*** .14281117 3.696 .0002 .3239437|
|ECONPOS4| .63638*** .16133022 3.945 .0001 .0845070|
|Constant| -1.81231*** .54906879 -3.301 .0010 |
|TREAT_PR| .18384*** .04980352 3.691 .0002 .4929577|
|TREAT_TR| -.02049 .03827036 -.535 .5923 .5070423|
|PER_PRIC| -.02757*** .00698740 -3.946 .0001 2.7112676|
+--------+------------------------------------------------------------+
| Note: ***, **, * = Significance at 1%, 5%, 10% level. |
+---------------------------------------------------------------------+
_______________________________________________
Limdep site list
Limdep at limdep.itls.usyd.edu.au
http://limdep.itls.usyd.edu.au
More information about the Limdep
mailing list