.-
help for ^ivprob^, ^divprob^, and ^ivtobit^                    
.-

Instrumental variables probit and tobit using Amemiya Generalized Least Squares -------------------------------------------------------------------------------

^ivprob^, ^divprob^, ^ivtobit^

^ivprob^ and ^ivtobit^ are used for estimating probit models where one or more of the independent variables is endogenous. Neither weights nor the @robust@ option are implemented.

^divprob^ produces marginal effects for ^ivprob^.

Description -----------

These programs implement Amemiya Generalized Least Squares (AGLS) estimators fo > r probit and tobit with endogenous regressors. Newey (1987, eq. 5.6) provides the formu > las used. The endogenous regressors are treated as linear functions of the instrum > ents and the other exogenous variables.

For a good summary of how AGLS works, see Maddala (1983) pp. 247-252.

Syntax ------

^ivprob^ depvar [^if^ exp] [^in^ range], endog(varlist1) iv(varlist2) exog(var > list3) ^divprob^ depvar [^if^ exp] [^in^ range], endog(varlist1) iv(varlist2) exog(va > rlist3) [CLAssic] ^ivtobit^ depvar [^if^ exp] [^in^ range], endog(varlist1) iv(varlist2) exog(va > rlist3) [ll(#) ul(#)]

where:

depvar is the dependent variable, varlist1 are the endogenous variables, varlist2 are the instruments for varlist1, varlist3 are the other exogenous variables, and ll and ul are the censoring points for the tobit (see @tobit@) with defaults > of zero (for ll) and 9999999 (for ul). For a description of the ^classic^ option with ^divprob^, see @dprobit@.

References ----------

Maddala, G. (1983). Limited-Dependent and Qualitative Variables in Econometric > s. Cambridge University Press.

Newey, W. (1987). "Simultaneous estimation of limited dependent variable model > s with endogenous explanatory variables." Journal of Econometrics, 36: 231-25 > 0.

Acknowledgements ----------------

These programs were inspired by Jonah Gelbach's -probitiv- and -tobitiv- (avail > able at http://glue.umd.edu/~gelbach/ado/). -ivprob- and -ivtobit- provide correct standard errors. They should also be more efficient than Gelbach's implementat > ion.

^divprob^ draws heavily on Stata's ^dprobit^.

Author -------

Joe Harkness, Johns Hopkins University, USA joe.harkness@@jhu.edu

Also see --------

@ivreg@ @treatreg@ (stb 55)