Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Marginal effects for first stage of hurdle model?

    I am running an exponential hurdle model with an indicator variable for treatment status. In the example below, I estimate the model & then use margins to get the marginal effect of treatment status in the second stage. What I would like to do is also report the first stage coefficient for treat (ie. the selection equation) so that the coefficient is interpretable. Is there a way to call margins on the first stage so that I can report either the marginal effect or the odds ratio?

    Code:
    webuse fitness
    set seed 123
    gen treat = runiformint(0,1)
    churdle exponential hours i.treat age i.smoke distance i.single, select(i.treat commute whours age) ll(0) nolog
    margins,dydx(i.treat)

  • #2
    Have a look at example 1 of the pdf documentation for -help churdle-. Margins will give you effects which are interpretable.

    Comment


    • #3
      Thanks Leonardo, I've consulted the help file & the link to postestimation but I haven't found anything on obtaining the first stage- I can obtain the estimate for the effect of treat on hours, what I am trying to export is an interpretable coefficient for hours in the selection estimation.

      Comment


      • #4
        Oh, I had misinterpreted your request. My apologies. In that case I don't think Stata will provide that to you. You could reconstruct the equation given the formulas specified and the printed coefficients from the selection model.

        Comment


        • #5
          The first step of churdle is just a probit model, so just use that:

          Code:
          gen hourspos = hours > 0
          replace hourspos = . if hours == .
          probit hourspos i.treat commute whours age
          margins, dydx(treat)

          Comment


          • #6
            Ah thank you Jeff- I had assumed for some reason it was estimating a logit and then I couldn't replicate the coefficient, that would explain why!

            Comment

            Working...
            X