[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: ML estimation of bivariate ordered probit models with endogeneity
I'd like to share with you the new program -bioprobit- : maximum-likelihood
estimation of bivariate ordered probit models. The setup is more ore less
straightforward generalization of univariate ordered probit (or bivariate
we have latent variables
where corr(e_1, e_2)=rho
observed are y_1=1...J, y_2=2...K (in fact program works with any coding of
dependent variables not just from 1 to J or K).
If -endogenous- option is used, a modified model
will be estimated ("standard" identification requirements apply).
-bioprobit- estimates all betas, rho, cutoff points and gamma. it is
implemented as a d2 procedure.
The program and an accompanying paper was submitted to Stata Journal. The
paper gives more detailed description of the model and also presents results
of Monte Carlo simulations. You can find a draft version at
Program can be installed by:
.net install bioprobit,
note: since Stata is case-sensitive you'll need to type the url as given
by typing :
. net from http://siteresources.worldbank.org/INTPOVRES/Resources
you'll find some programs written by us on poverty and inequality analysis,
and other useful tools, including -xml_tab- module to export Stata output to
a printer-ready formatted tables in an xml file directly openable in MS
Excel and OpenOffice.org Calc;
recent addition to the site is program -bestreg- that implements the Leaps
and Bounds algorithm for best subset selection for the regression analysis.
It's still in the beta version but since there was some interest in tools
for all possible regressions on Statalist recently I decided to put it
online as well.
And of course, I'd appreciate any comment or suggestion.
* For searches and help try: