[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: predict and adjust after reg3
I have been searching for a less cumbersome way (other than doing this
one variable at a time, recoding the variable, reinstating it, and
changing another variable) that would do what adjust or the prchange
program of Scott Long or Gary King's clarify would do for single
Specifically I have a multiple equation model estimated by reg3,
reg3 (y1 y2 x1 x2 x3 ) (y2 y3 y4 x3 x4 x5 x6 ) (y3 y1 x5 x7 x8 x9)
and have computed predicted values for each equation using predict.
Now what I want to do is to get predictions based upon the minimum and
maximum values of the predictors (y2 x1 x2 x3 --respectively, not all at
the same time--for the equation for y1, etc. for the equations for y2
So what I am looking for is: what happens to y1 if all cases
hypothetically had the value on y2 of y2(minimum) and did not change on
x1 x2 and x3).
And what would happen to y1 if all cases hypothetically had the value
on y2 of y2(maximum) and did not change on x1 x2 and x3).
E.g., a specific example is that in a model for corruption using trust
as a(n) (endogenous) predictor, what would the predicted level of
corruption be if a country "became" as trusting as Norway or as
mistrusting as Brazil (maximum and minimum values, respectively),
holding the other predictors constant?
I've done this before, but I have now reestimated a bunch of models and
the previous efforts were one by one and took a ton of time.
You can get this from adjust, prchange (for probit/logit), and clarify
(estsimp) but not for multiple equation models.
Many thanks for any help,
Department of Government and Politics
University of Maryland
College Park, MD 20742
office phone: 301-405-4151
home phone 301-279-0414
office fax: 301-314-9690
home fax: 301-279-2614
home page: http://www.bsos.umd.edu/gvpt/uslaner
new page on corruption research:
* For searches and help try: