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

  • (I NEED HELP) How to manually compute generalized residuals and score for a right-censored Tobit model?

    Hello, Statalist fellows!

    I'm finishing my bachelor's thesis, in which I'm running a right-censored Tobit model for a (0, 1] dependant variable.

    After the regression, I need to compute the generalized residuals to check for normality and homoskedasticity. Sadly, the only formal reference that I got is from Cameron and Trivedi (2010) "Microeconometrics using Stata" and it is for a log transformed dependant variable and a left-censored estimation. I have severe doubts about computing the inverse Mills ratio, especially when I generate the threshold.

    Please, help me.

    Thanks in advance.


  • #2

    As my dependent variable is right-censored at 1, I generated the residuals

    predict pvdepvar, ystar(.,1). (predicted values of dependant variable, right-censored at 1).
    generate residuals= depvar-pvdepvar

    My residuals had good kurtosis and are left skewed, but they passed (barely) the sktest

    . summarize residuals

    Variable | Obs Mean Std. Dev. Min Max
    residuals | 95 -.0018706 .1035449 -.2949046 .190228

    . sktest residuals

    Skewness/Kurtosis tests for Normality
    ------ joint ------
    Variable | Obs Pr(Skewness) Pr(Kurtosis) adj chi2(2) Prob>chi2
    ------------- +-------------------------------------------------------------------------------
    residuals | 95 0.0152 0.5798 5.91 0.0521

    I hope this helps others!
    Last edited by Romar Guerra; 19 Jan 2020, 08:23.