Announcement

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

  • FE RE Model

    Hello everyone. I wanted to know apart from testing which model to use , fixed effects or random effects through Hausman test on Stata, how can I provide theoretical justification for using FE or RE model. In my case I have 30 manufacturing firms and time=20 years

  • #2
    If you think that in the model

    Yit = b*Xit + Ui + Vit

    Ui is correlated with (some of) the regressors Xit, then you do fixed effects.

    Otherwise random effects.

    Comment


    • #3
      Thank Joro. But how do I determine if I have to include a Ui in the model. That's what am confused about.

      Comment


      • #4
        I presume you mean how to decide whether Ui is correlated with the regressors or not?

        You determine it from economic knowledge of the situation. You should have some notion of what goes into Ui and what goes into Xit, and you should have some notion of whether these are correlated or not.

        The second statistical way is what you mentioned, by Hausman test. The Hausman test estimates the b from the FE model, and the b from the RE model, and then tests whether the two estimated b are the same. If they are not, we go with the FE model.

        Originally posted by Anuradha Saikia View Post
        Thank Joro. But how do I determine if I have to include a Ui in the model. That's what am confused about.

        Comment


        • #5
          Thank you Joro

          Comment

          Working...
          X