Announcement

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

  • nehurdle available on SSC

    Hi everybody,

    I would like to announce that thanks to Kit Baum, nehurdle is now available at SSC.

    nehurdle is a command to estimate models where the dependent variable has a corner solution at 0 (see chapter 17 of Wooldridge (2010)) via maximum likelihood. It collects the three more popular models for such data: Tobit, Truncated hurdle, and Type II Tobit. It also allows for a linear or exponential specification of the value equation in all three models. Furthermore, it allows to model multiplicative heteroskedasticity in the value and selection equations in all three models (remember that in a Tobit model the selection and value equations are one and the same). The command also has its own predict functionality that makes prediction of the censored variable very easy, something that is crucial for these types of variables since the censored variable is the observed variable.

    I have written an article that illustrates some of its functionalities and will be published in the Stata Journal in the near future. I hope the help files in the meantime allow you to work with it. Please let me know if you have any questions about the command.

    Reference:
    Wooldridge, Jeffery M. 2010. Econometric Analysis of Cross Section and Panel Data. 2nd ed. Cambridge, MA: MIT Press.
    Alfonso Sanchez-Penalver

  • #2
    Hello Alfonso,

    I am using the command nehurdle on stata and was wondering what type of pseudo R-squared is used for the output? Is it a McFadden’s pseudo R-squared?

    Thank you in advance

    Comment


    • #3
      Hi Hugo, -nehurdle- predicts the censored mean for all observations and reports the squared Pearson correlation coefficient of the censored mean and the actual observed variable as the pseudo R-squared. You can get to it by typing
      Code:
      predict double ycen
      correl y ycen
      display `r(rho)'^2
      where y is your dependent variable, after having run -nehurdle-.
      Alfonso Sanchez-Penalver

      Comment


      • #4
        Thank you so much Alfonso. How could this number therefore be interpreted? Thank you again for your help, really much appreciated.

        Comment


        • #5
          As an R-squared. For the usual linear model it is identical to the R-squared that is reported. It is the percentage of the variation of the dependent variable that is explained by the variation in the independent variables. Notice that when you are modeling an exponential value function, it will still represent the share of the variation dependent variable (the one in levels not in logs), even though the transformation to logs was done to estimate the model. So it is always the percentage of the variation of the dependent variable as you observe it that is explained by the model.
          Alfonso Sanchez-Penalver

          Comment


          • #6
            By the way, feel free to let me know if you like the command!!!
            Alfonso Sanchez-Penalver

            Comment


            • #7
              Thank you again, you've been of great help. The command works really well and it's helping me for my report, thank you!

              Comment


              • #8
                Glad to hear that. Best of luck on your report.
                Alfonso Sanchez-Penalver

                Comment


                • #9
                  Hi Alfonso.

                  Greetings.

                  Can you please share the command/syntax and any related documents.

                  Kind Regards,

                  Samir

                  Comment


                  • #10
                    Hi Samir,

                    to access the package type ssc install nehurdle in your Stata command prompt. You will need an internet connection, and the software will download the command program and help files.

                    If you want some examples and discussion about the command, here is the link to the article in The Stata Journal

                    https://journals.sagepub.com/doi/abs...36867X19830893

                    Best,

                    Alfonso
                    Alfonso Sanchez-Penalver

                    Comment


                    • #11
                      Prof Alfonso Sánchez-Peñalver Please suggest if it is possible to implement the lognormal hurdle model instead of the truncated normal hurdle with your command? I am using Stata12. Thank you.

                      Comment


                      • #12
                        Hi Parul, as explained in the help you can estimate the natural logarithm of the value function for any of the three specifications, Tobit, Truncated Hurdle and Type II Tobit, by simply adding the exponential option.
                        Alfonso Sanchez-Penalver

                        Comment

                        Working...
                        X