Announcement

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

  • heteroskedasticity problem ‎in 3SLS and CMP command erorr

    Dear Statalist

    I have a heteroskedasticity problem in my 3SLS model. When I am using CMP from SSC in Stata 13, the following error is reported to me:

    "Warning: regressor matrix for m1 equation appears ill-conditioned. (Condition number = 21.37175.)
    This might prevent convergence. If it does, and if you have not done so already, you may need to remove nearly collinear regressors to achieve convergence. Or you may need to add a nrtolerance(#) or nonrtolerance option to the command line." same as this errors is reported for M2 and M3

    also my model is:

    cmp( m1= m2# m3# debtequity roe tangibleasset)( m2= m1# m3# debtequity roe exe)( m3= m1# m2# debtequity roe interesttaxshield currentratio ), ind ($cmp_cont $cmp_cont $cmp_cont ) nolr tech(dfp) qui


    I could not run the nrtolerance(#) in CMP. May I ask you how should I use it? could you give me a sample command for nrtolerance(#)?

    Moreover, I have heteroskedasticity problem when I ran the 2SLS for above model. Unfortunately, I do not know any other robust model for heteroskedasticity problem in 3SLS and 2sls. May I kindly know if you suggest another robust model for 3SLS and 2SLS?

    Best regards
    maziar

  • #2
    "#" stands for a number, like .001. Type "help maximize" to learn more about this option. But notice the message you got is only a warning and suggests you try this option only if your model fails to converge without it.

    Comment


    • #3



      Dear Roodman,
      thank you for your comment. I ran the nrtolerance(#) with different value. But the model could not run. May I ask you have a look to the following Stata results? I put the results of reg3, nrtol and nonrtol. also I checked the nrtol with different values all the results were same as each other. There is may be a fault that I could not understand.thank you so much for your consideration.


      Three-stage least-squares regression
      ----------------------------------------------------------------------
      Equation Obs Parms RMSE "R-sq" chi2 P
      ----------------------------------------------------------------------
      m1 2241 5 16.12645 -0.2955 99.47 0.0000
      m2 2241 5 2.211189 -0.1290 137.23 0.0000
      m3 2241 6 .2307967 -0.5483 339.94 0.0000
      ----------------------------------------------------------------------

      -----------------------------------------------------------------------------------
      | Coef. Std. Err. z P>|z| [95% Conf. Interval]
      ------------------+----------------------------------------------------------------
      m1 |
      m2 | 4.220506 .779838 5.41 0.000 2.692052 5.74896
      m3 | 6.511411 4.201444 1.55 0.121 -1.723269 14.74609
      ldiv | -.9892566 .1462237 -6.77 0.000 -1.27585 -.7026635
      roe | .0920138 .0405287 2.27 0.023 .0125791 .1714485
      tangibleasset | 5.842517 1.794619 3.26 0.001 2.325129 9.359905
      _cons | -45.33092 10.20529 -4.44 0.000 -65.33293 -25.32891
      ------------------+----------------------------------------------------------------
      m2 |
      m1 | .0693731 .0274738 2.53 0.012 .0155255 .1232208
      m3 | -.0482026 .6296454 -0.08 0.939 -1.282285 1.18588
      roe | -.0170742 .0052655 -3.24 0.001 -.0273944 -.0067541
      ldiv | .1325258 .0239901 5.52 0.000 .085506 .1795456
      exe | .2102599 .047029 4.47 0.000 .1180847 .302435
      _cons | 11.83257 .3184304 37.16 0.000 11.20846 12.45668
      ------------------+----------------------------------------------------------------
      m3 |
      m1 | -.0082954 .0045497 -1.82 0.068 -.0172127 .0006219
      m2 | .0652459 .0249859 2.61 0.009 .0162745 .1142173
      ldiv | -.0123001 .0052012 -2.36 0.018 -.0224942 -.0021059
      roe | -.0007775 .0007008 -1.11 0.267 -.0021512 .0005961
      interesttaxshield | -.1712144 .0365336 -4.69 0.000 -.242819 -.0996098
      currentratio | -.0198191 .0012939 -15.32 0.000 -.022355 -.0172831
      _cons | -.4260373 .274067 -1.55 0.120 -.9631988 .1111242
      -----------------------------------------------------------------------------------



      cmp( m1= m2# m3# ldiv roe tangibleasset)( m2= m1# m3# roe ldiv exe)( m3= m1# m2# ldiv roe interesttaxshield currentratio ), ind( $cmp_cont $cmp_co
      > nt $cmp_cont) nonrtol tech(dfp) qui
      (27 observations dropped from m1 equation because they are unavailable in the m2 equation, on which the m1 equation depends)
      (131 observations dropped from m1 equation because they are unavailable in the m3 equation, on which the m1 equation depends)
      (131 observations dropped from m2 equation because they are unavailable in the m1 equation, on which the m2 equation depends)
      (27 observations dropped from m3 equation because they are unavailable in the m1 equation, on which the m3 equation depends)
      Fitting individual models as starting point for full model fit.
      Note: For programming reasons, these initial estimates may deviate from your specification.
      For exact fits of each equation alone, run cmp separately on each.
      Fitting constant-only model for LR test of overall model fit.
      Fitting full model.
      Mixed-process regression Number of obs = 2241
      LR chi2(8) = -18045.99
      Log likelihood = -22361.834 Prob > chi2 = 1.0000
      -----------------------------------------------------------------------------------
      | Coef. Std. Err. z P>|z| [95% Conf. Interval]
      ------------------+----------------------------------------------------------------
      m1 |
      m2 | 11.63696 .6355415 18.31 0.000 10.39132 12.8826
      m3 | -17.26419 . . . . .
      ldiv | -2.591297 . . . . .
      roe | 45.93337 .8521784 53.90 0.000 44.26313 47.60361
      tangibleasset | -5.580363 3.681792 -1.52 0.130 -12.79654 1.635817
      _cons | -70.12161 . . . . .
      ------------------+----------------------------------------------------------------
      m2 |
      m1 | -1.202589 . . . . .
      m3 | 67.61488 2.509017 26.95 0.000 62.69729 72.53246
      roe | -24.89055 . . . . .
      ldiv | 1.642289 .1376361 11.93 0.000 1.372527 1.912051
      exe | -1.776823 .250474 -7.09 0.000 -2.267743 -1.285903
      _cons | -55.70998 5.699915 -9.77 0.000 -66.88161 -44.53835
      ------------------+----------------------------------------------------------------
      m3 |
      m1 | -185.7407 6.951277 -26.72 0.000 -199.3649 -172.1164
      m2 | -104.4102 . . . . .
      ldiv | 59.81647 . . . . .
      roe | -8.004291 34.9612 -0.23 0.819 -76.52698 60.5184
      interesttaxshield | .2291701 313.591 0.00 0.999 -614.3979 614.8563
      currentratio | 6.042624 11.11328 0.54 0.587 -15.739 27.82425


      cmp( m1= m2# m3# ldiv roe tangibleasset)( m2= m1# m3# roe ldiv exe)( m3= m1# m2# ldiv roe interesttaxshield currentratio ), ind( $cmp_cont $cmp_co
      > nt $cmp_cont) nrtol(.00001) tech(dfp) qui
      (27 observations dropped from m1 equation because they are unavailable in the m2 equation, on which the m1 equation depends)
      (131 observations dropped from m1 equation because they are unavailable in the m3 equation, on which the m1 equation depends)
      (131 observations dropped from m2 equation because they are unavailable in the m1 equation, on which the m2 equation depends)
      (27 observations dropped from m3 equation because they are unavailable in the m1 equation, on which the m3 equation depends)
      Fitting individual models as starting point for full model fit.
      Note: For programming reasons, these initial estimates may deviate from your specification.
      For exact fits of each equation alone, run cmp separately on each.
      Fitting constant-only model for LR test of overall model fit.

      Fitting full model.
      cannot compute an improvement -- flat region encountered
      convergence not achieved
      convergence not achieved
      Last edited by maziar ghasemi; 12 Sep 2015, 08:36.

      Comment

      Working...
      X