I'm working with a dependent variable $y_i \in [0,1]$. I have a single endogenous explanatory variable $w_i \in [0, 1]$ with corresponding instrument $z_i \in \{0,1\}$. Suppose further that I have a vector of exogenous explanatory control variables $\vec{x_i}$.

My intuition is to estimate a two-stage tobit model. Is this possible? Are there any caveats to doing this? Can I do this in R?

  • $\begingroup$ Does the zero correspond to censoring or a corner solution? $\endgroup$
    – dimitriy
    Dec 5, 2014 at 3:01
  • $\begingroup$ Sorry, I should have added that the zero corresponds to a corner solution in both cases. $\endgroup$ Dec 5, 2014 at 3:20
  • $\begingroup$ Correction: the [0,100] closed interval corresponds to two corner solutions. $\endgroup$ Dec 5, 2014 at 13:12
  • $\begingroup$ You might find my answer to this question useful. stats.stackexchange.com/questions/66856/… $\endgroup$
    – Bill
    Dec 5, 2014 at 17:29


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy