st: using -mfx- or -margins- after the Wooldridge approach

[Igor Ramzak <igor.ramzak@gmail.com>]

Dear statalisters,

I estimate Wooldridge's Conditional Maximum Likelihood for a Dynamic
Probit Model.
By using the -xtprobit- command, I regress having a panic attack in
period t on several independent variables, the longitudinally averaged
variables for each person (using -bysort personid: egen var1_mean =
mean(var1)- ) and the initial outcome of the dependent variable.

In his paper (Wooldridge 2005), chapter 4 explains how to calculate
the average partial effects. Can anybody help me to implement this in
Stata?
I use -mfx compute, predict (pu0)- but I'm not sure this is correct.

A supplimentary problem... Some of my variables are mutually
exclusive. For example, the type of commuting to work is one group,
captured by 2 dummy variables:
* publictransport
* car
* by foot or by bike as reference group
For example, I want to know the marginal effect of using public
transport versus going by foot/bike. Do I have to take this in account
somehow? When done automatically, the mean value of the dummy variable
"car" is probably used. Should it be fixed to zero? How can I do that
using -mfx- or -margins-?

Thanks for your consideration.
Igor


References:

Wooldridge, 2005, "Simple solutions to the initial conditions problem
in dynamic, nonlinear panel data models with unobserved
heterogeneity", Journal of Applied Econometrics n°20, p39-54.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/