# st: Simulating a change in a variable and obtaining its effects on the predicted probabilities of a multinomial probit.

 From Stephen Johnston To majordomo@hsphsun2.harvard.edu Subject st: Simulating a change in a variable and obtaining its effects on the predicted probabilities of a multinomial probit. Date Tue, 19 Sep 2006 16:25:17 -0400

Hello,

I am trying to figure out how to simulate the effect of a change of a variable on the predicted probabilities of a multinomial probit.

For simple probit model you can do as follows:

probit y x1 x2 x3
gen beta_x1=_b[x1]
gen beta_x2=_b[x2]
gen beta_x3=_b[x3]
gen cons =_b[_cons]

gen x1b=beta_x1*x1
gen x2b=beta_x2*x2
gen x3b=beta_x3*x3

gen probability=normal(x1b+x2b+x3b+cons)

This gives the same results as the 'predict' command after a probit.

Then, to simulate a rise of 1 unit in x2 for example, I think you can do as follows:

gen new_x2=x2+1
gen new_x2b=beta_x2*new_x2

and then to find the new predicted probabilities,

gen prob_new=normal(x1b+new_x2b+x3b+cons)

Does anyone know how to perform a similar procedure for the mprobit command?

Thanks,
Steve
*
* 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/