Announcement

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

  • Calculate the occurrence probability of crisis in the probit model

    Dear Statalist members,
    I want to get the probability of a crisis after estimating a probit model using panel data. Given the complexity of the probit function, how do I calculate the probability of occurrence of one of the states (the crisis) in binary probit regression of panel data?

    Thanks in advance

  • #2
    Mohammad: Can you show us the code you use to estimate your model?

    Comment


    • #3
      In the pdf documentation provided with Stata, the chapter on every or almost every Stata estimator is followed by a chapter titled "estimator postestimation". Have you looked at xtlogit postestimation in the documentation?

      Comment


      • #4
        Hi John
        the code for the estimate of the probit model is:

        xtprobit crisis0 l.drer l.gdc l.ggdp l.inf l.irid l.ld l.m2d l.m2ir l.rus l.xir l.xm, re

        of course, the codes for estimate of logit model and calculate the occurrence probability of crisis in my research are:

        xtlogit crisis0 l.drer l.gdc l.ggdp l.inf l.irid l.ld l.m2d l.m2ir l.rus l.xir l.xm, re
        predict crisishat
        generate crisis_p = 1/(1+ exp(-crisishat))
        replace crisis_p = 1 if crisis_p > 0.1
        replace crisis_p = 0 if crisis_p <= 0.1
        tab crisis0 crisis_p

        Thanks in advance

        Comment


        • #5
          Hi Phil,
          Yes, I did and I don't think it exists in the postestimation. Of course, as I mentioned in the above, I've got these probabilities for xtlogit using the above codes, but I have not been able to do this with xtprobit.
          Thank you if you can help

          Comment


          • #6
            Mohammad: After your command
            Code:
            xtprobit crisis0 l.drer l.gdc l.ggdp l.inf l.irid l.ld l.m2d l.m2ir l.rus l.xir l.xm, re
            did you try
            Code:
            predict predcrisis0, pr

            Comment


            • #7
              Hi John,
              As I mentioned in 4 post above, the result of this code in the Logit model, is not the probability of a crisis, and we have to put this result into another relationship to calculate the probability of a crisis. Now, my question is how can we convert the predicted value of this code into the probability of a crisis in the probit model?

              Comment


              • #8
                Hi Mohammad hope you've managed to find a solution...and perhaps you or someone can help me...I want to run a panel logit crisis model and not sure how to code the crisis dependent variable that requires: 1, non-performing assets/asset >10%, bank bail out >2% of GDP, emergency govt intervention and bank runs...ll appreciate any help really...thanks in advance.

                Comment

                Working...
                X