 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: new estimation command for mixed, multi-eq probit, tobit, continuous
I'm happy to announce a new estimation command. It is called "cmp," for
conditional mixed process. It fits multi-equation models that can mix
probit, tobit, and "continuous" (unbounded/OLS-like) dependent
variables. In effect, it subsumes the official commands biprobit,
ivprobit, and ivtobit, and the user-written mvprobit, bitobit,
triprobit, and mvtobit, and does much more. The "conditional" means that
which equations are included in the model can vary by observation. The
program requires Stata 10.
So for example, you could a regress a censored variable on two
endogenous variables, one binary and one continuous and uncensored, each
instrumented with other variables.
To install, type "ssc install cmp". This is brand new, so I welcome
comments.
A slightly more detailed explanation, from the help file:
cmp estimates multi-equation, conditional recursive mixed process
models. "Mixed process" means that different equations can have
different kinds of dependent variables. The choices are: continuous
(like OLS), tobit (left-, right-, or bi-censored), and probit. A
dependent variable in one equation can appear on the right side of
another equation. However, cmp can only fit "recursive" models with
clearly defined stages, not ones with simultaneous causation. A and B
can be determinants of C and C a determinant of D--but D cannot be a
determinant of A, B, or C. And the "conditional" means that the model
can vary by observation. An equation can be dropped for observations for
which it is not relevant--if, say, a worker retraining program is not
offered in a city then the determinants of uptake cannot be modeled
there. Or the type of dependent variable can vary by observation, though
the utility of this freedom is less obvious
cmp's modeling framework therefore embraces those of the official Stata
commands probit, ivprobit, biprobit, tobit, and ivtobit, in principle
even regress, sureg, ivreg, as well as the user-written triprobit,
mvprobit, bitobit, and and mvtobit. It goes beyond them in offering far
more flexibility in model construction. To take one arbitrary example,
one could regress a continuous variable on two endogenous variables, one
binary and the other sometimes left-censored, instrumenting each with
additional variables. And it allows the models to vary by observation.
In some cases, the gain is consistent estimation where it was difficult
before. In other cases, the gain is just in efficiency. For example if y
is continuous, x is a sometimes-left-censored determinant of y, and z is
an instrument, then the effect of x on y can be consistently estimated
with 2SLS (Kelejian 1971). However, a cmp estimate that uses the
information that x is censored will be more efficient, based as it is on
a more accurate model.
David Roodman
Research Fellow
Center for Global Development
[email protected]
http://www.cgdev.org
+1 (202) 416-0723
1776 Massachusetts Ave. NW
Washington, DC 20036
*
* 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/