Announcement

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

  • FE (spatial) with lagged variables

    Hi everyone

    I am using Spatial autoregressive model for panel data - spatial fixed effect model (spxtregress). I need my independent variables to be time-lagged. This is not applicable directly when spxtregress command is used. Hence, instead, I create lagged variables in the following manner

    gen lx=l.x
    gen lx1=l.x1
    gen lx2=l.x2
    I use these variables within the following command

    Code:
    spxtregress  Y lx lx2 lx3, fe  dvarlag(N) ivarlag(N: lx lx2 lx3) errorlag(N)
    N is the spatial weight matrix. However the output I get is the following:

    Code:
    Fixed-effects spatial regression                Number of obs     =        264
    Group variable: _ID                             Number of groups  =         33
                                                    Obs per group     =          8
    
                                                    Wald chi2(4)      =     121.89
                                                    Prob > chi2       =     0.0000
    Log likelihood =   173.6341                     Pseudo R2         =     0.0972
    
    ------------------------------------------------------------------------------
             Y   |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
    -------------+----------------------------------------------------------------
    Y            |
     lx         |          0  (omitted)
      lx1        |          0  (omitted)
      lx2         |          0  (omitted)
    -------------+----------------------------------------------------------------
    N            |
     lx           |   2.899334   1.527289     1.90   0.058    -.0940967    5.892765
      lx1         | |  -6.070934    2.52357    -2.41   0.016    -11.01704   -1.124828
       lx2          |   2.095667   1.374309     1.52   0.127     -.597929    4.789264
             Y   |   .6944136   .1342155     5.17   0.000      .431356    .9574712
           e.Y |  -.4444192   .4677527    -0.95   0.342    -1.361198    .4723593
    -------------+----------------------------------------------------------------
        /sigma_e |    .111479   .0055633                      .1010914    .1229339
    ------------------------------------------------------------------------------
    Wald test of spatial terms:          chi2(5) = 165.28     Prob > chi2 = 0.0000
    Can someone explain to my why I am getting omitted variables lx-lx2 in the results output, while spatial effects are still being calculated? I would appreciate if someone can give a tip how to circumvent this issue.
    I also used standard FE model with xtreg command instead. There was no problem- lx-lx2 were not omitted.

    Thank you in advance!
    Mina Wu


  • #2
    Hi,
    I am having a similar issue. Did you ever find a solution to your problem?

    Comment


    • #3
      I am also having the same issue

      Attached Files

      Comment


      • #4
        -spxtregress- is designed to fit static panel-data spatial autoregressive models, so we cannot add the time-lagged independent variables or dependent variables in -spxtregress-. The -estat impact- also assumes static models in the computations of impacts. This is true for both fixed-effects and random-effects estimator. If you add a time-lagged variable, no matter it is expressed as L.x or another new variable, the inference is no longer valid.

        Comment


        • #5
          I am also facing same problem.

          Comment

          Working...
          X