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

  • Heckman Selection Bias

    Hello everyone,

    I am writing you because I am trying to run a heckman regression for testing selection bias and I have a doubt.

    Is it Heckman exclusive for censored variables? I have this question because I was trying to run the regression and I had a message "Dependent variable never censored because of selection:
    model would simplify to OLS regression", then I checked the stata help and I found that an example about wage include some missing values (instead of of 0s) in the dependent variable, then I create a new variable transforming 0s in missing values. The regression Works.

    My dependent variable is a proportion, that takes values between 0 and 0.86 but is not censored, naturally has 0 values. The sample is N=25000 and about 8500 are 0.

    Finally, I would like to ask you how do I interpret the mills lambda?

    Thank you very much for your kind help.

    Best regards,

    Last edited by Alejandro Torres; 11 Jul 2018, 18:00.

  • #2
    You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

    Heckman assumes you have observations with an observed dv, and observations without the observed dv. If you don't have this, there is no way to estimate the selection part of the model. It looks like you don't have any missing values on the dv. That's why converting the 0's into missing let you estimate the heckman. However, if you have natural zeros (legitimate values), then treating them as missing is not advisable. You may want to look at the tobit estimators which are designed for such problems.


    • #3
      Thank you very much Phil, I really appreciate your time answering.
      Best reagards,