Announcement

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

  • Which panel model should i choice when using the lag of independent variable.

    Dear Stata members,

    I have a question about using lag of independent variables in my panel data model.
    This is the panel model I will use: Yit=a + b*Xit-1 + c*Zit-1+uit
    According to the theoretical model I need to take one lag of independent variables. Because of not taking lag of dependent variable (Yit), I will not use GMM or ARDL model.
    Can I use static panel data model (POLS, FE, RE) to predict this model? Is there any problem if I use static panel data model?
    Thank you very much.

  • #2
    As long as the regressors you employ are exogenous and theory dictates that you should employ lagged regressors, to the best of my knowledge you should be fine. I have searched and have found no literature suggesting this practice is wrong. Perhaps someone will correct me though.

    Comment


    • #3
      Dear Morlet,

      Thank you for your answer.
      In the literature of my work, researcher used a lag of independent variables and analyzed that with POLS model. But this literature was a bit old. Made between 1975-2010.
      Maybe I should use "Distributed Lag Models (DL)". Now I'm searching DL Model.

      Comment


      • #4
        If I were you, now I would go about determining the appropriate model to use. POLS is likely to yield biased results. Conduct a BPLM test (xttest0), then conduct a test for groupwise heteroscedasticity and within group autocorrelation. If both or even one of these tests shows heteroscedasticity or autocorrelation, do not use a Hausman (1978) test to discriminate between fixed and random effects, but a Mundlak (1978) test described by Enrique Pinzon here: https://blog.stata.com/2015/10/29/fi...dlak-approach/.

        Then you would have to decide which standard errors are most appropriate. Further test for the normal distribution of your residuals and their cross-sectional correlation.

        Comment


        • #5
          Thank you very much for your help.

          Comment

          Working...
          X