[Limdep Nlogit List] Ordered probit model with selectivity-partial effects
b.quinn at qub.ac.uk
Tue Jul 13 03:01:21 EST 2010
I am estimated the above model using the command below
I would like to calculate partial effects for this model (ideally APEs, but at the mean of the observations would be a start), and the standard command (;margin) doesn't produce an output.
I therefore would like to do this manually using matrix commands and have a few questions re how to proceed
I would like to calculate both the direct(from index probability function) and indirect(from selection equation)
-For direct effects of each outcome I should use only the subsample(ie y=1) that is used to estimate this model?
-For indirect effects I should use the full sample?
-If anyone had some code re the calculation of average partial effects for an ordered probability I would be most grateful and hopefully I could adapt it to the case of selectivity.
I have tried and failed with
I also used the E31.2.6 example in the manual which uses simulated data and still no margial effects using ;margin$.
From: limdep-bounces at limdep.itls.usyd.edu.au [mailto:limdep-bounces at limdep.itls.usyd.edu.au] On Behalf Of Liren Zhou
Sent: 12 July 2010 05:26
To: limdep at limdep.itls.usyd.edu.au
Subject: [Limdep Nlogit List] Parameter constrain
I am trying to estimate a simple MNL model using following code. I need
to constrain coefficients for some variables. For example, set the
coefficient of travel time to 0.07. Can anybody tell me how it works?
Thanks a lot in advance!
NLOGIT; LHS = P_HBO, CASENUM, P_INDEX
; CHOICES = AM, MD, PM, NT
; MODEL: U(AM) =Time*Time_HBO/
U(MD) = MD_C + MD_SR3*MODE_SR3 + Time*TIME_HBO + MD_SIZE*HH_SIZE/
U(PM) = PM_C + PM_SR3*MODE_SR3 + Time*TIME_HBO + PM1_SIZE*HH_SIZE/
U(NT) = NT_C + NT_SR3*MODE_SR3 + Time*Time_HBO + NT_SIZE*HH_SIZE$
Limdep site list
Limdep at limdep.itls.usyd.edu.au
More information about the Limdep