Announcement

Collapse
No announcement yet.
X
Collapse
  • Page of 1
  • Filter
  • Time
  • Show
Clear All
new posts
  • #1

    Poisson rather than the negative binomial/truncated Poisson rather than an untruncated Poisson

    Hello guys,

    1- How can I justify the use of the Poisson rather than the negative binomial with an econometric test in Stata?
    2- How can I justify the use of a truncated Poisson (without 0's) rather than an untruncated Poisson (with 0's) with an econometric test in Stata?

    Thank you,
    Tags: None
  • #2
    Hi,

    I tried this command

    Code:
    lrtest nbreg poisson
    I got this message:
    df(unrestricted) = df(restricted) = 17

    Is there someone who can help me with this?

    Thanks,

    Comment

    • #3
      I run this
      Code:
      . lrtest nbreg poisson, force
      I got this

      HTML Code:
      Likelihood-ratio test                                 LR chi2(0)  =   -832.43
(Assumption: poisson nested in nbreg)                 Prob > chi2 =         .
      Code:
      lrtest Poisson1 nbreg, force
      HTML Code:
      Likelihood-ratio test                                 LR chi2(0)  =      0.00
(Assumption: nbreg nested in Poisson1)                Prob > chi2 =         .

      Is the result acceptable?

      Comment

      • #4
        Dear Hm Saleh,

        Please explain to us why you are doing this because otherwise it is difficult to make sense of what you are trying to do.

        Best wishes,

        Joao
        • 1 like

        Comment

        • #5
          Dear @Joao Santos Silva

          I want to prove that Poisson is a suitable model for my data than Negative binomial regression.

          HTML Code:
          [CODE]nbreg TNCHILDREN i.Marital_Status Femaleearningsshareofhouseh HOUSEHOLD_INCOME  AGE age2 age3 ib1.HISPANIC HISPANIC_BORN_OUT_THE_US i.White_NON o.Asian_N
> ON i.Other_NON i.Black_NON i.EDUCATION[/CODE]. nbreg TNCHILDREN i.Marital_Status Femaleearningsshareofhouseh HOUSEHOLD_INCOME  AGE age2 age3 ib1.HISPANIC HISPANIC_BORN_OUT_THE_US i.White_NON o.Asian_N
> ON i.Other_NON i.Black_NON i.EDUCATION

Fitting Poisson model:

Iteration 0:   log likelihood = -3131.7546  
Iteration 1:   log likelihood = -3131.7533  
Iteration 2:   log likelihood = -3131.7533  

Fitting constant-only model:

Iteration 0:   log likelihood = -4094.3542  
Iteration 1:   log likelihood = -3204.6718  
Iteration 2:   log likelihood = -3204.6718  

Fitting full model:

Iteration 0:   log likelihood = -3132.4227  
Iteration 1:   log likelihood = -3131.7534  
Iteration 2:   log likelihood = -3131.7533  

Negative binomial regression                      Number of obs   =       2125
                                                  LR chi2(16)     =     145.84
Dispersion     = mean                             Prob > chi2     =     0.0000
Log likelihood = -3131.7533                       Pseudo R2       =     0.0228

---------------------------------------------------------------------------------------------
                 TNCHILDREN |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
----------------------------+----------------------------------------------------------------
             Marital_Status |
                         2  |  -.1570887   .0492608    -3.19   0.001    -.2536381   -.0605394
                         3  |  -.0438584   .0438118    -1.00   0.317     -.129728    .0420112
                            |
Femaleearningsshareofhouseh |  -.1502462   .0679401    -2.21   0.027    -.2834063   -.0170861
           HOUSEHOLD_INCOME |  -.0192837    .006056    -3.18   0.001    -.0311532   -.0074142
                        AGE |   .1685075   .1753308     0.96   0.337    -.1751345    .5121494
                       age2 |     -.0033   .0054151    -0.61   0.542    -.0139134    .0073133
                       age3 |   .0000218   .0000546     0.40   0.690    -.0000853    .0001288
                 2.HISPANIC |  -.1572284   .0915518    -1.72   0.086    -.3366667    .0222099
   HISPANIC_BORN_OUT_THE_US |  -.0766749   .0769876    -1.00   0.319    -.2275679    .0742181
                1.White_NON |   .0834247    .075069     1.11   0.266    -.0637079    .2305574
                  Asian_NON |          0  (omitted)
                1.Other_NON |   .2261337   .1080903     2.09   0.036     .0142807    .4379868
                1.Black_NON |   .2607497   .0855347     3.05   0.002     .0931047    .4283947
                            |
                  EDUCATION |
                         2  |   .0327911   .0472893     0.69   0.488    -.0598943    .1254765
                         3  |   .2063799    .068088     3.03   0.002     .0729299    .3398299
                         4  |   -.031971   .0584246    -0.55   0.584    -.1464812    .0825392
                         5  |   .0395897   .0448245     0.88   0.377    -.0482647    .1274441
                            |
                      _cons |  -1.690558   1.855493    -0.91   0.362    -5.327256    1.946141
----------------------------+----------------------------------------------------------------
                   /lnalpha |  -28.50159          .                             .           .
----------------------------+----------------------------------------------------------------
                      alpha |   4.19e-13          .                             .           .
---------------------------------------------------------------------------------------------
Likelihood-ratio test of alpha=0:  chibar2(01) =    0.00 Prob>=chibar2 = 1.000
          Code:
           tpoisson TNCHILDREN i.Marital_Status Femaleearningsshareofhouseh HOUSEHOLD_INCOME  AGE age2 age3 ib1.HISPANIC HISPANIC_BORN_OUT_THE_US i.White_NON o.Asia
> n_NON i.Other_NON i.Black_NON i.EDUCATION
          HTML Code:
          Iteration 0:   log likelihood = -2809.6753  
Iteration 1:   log likelihood = -2717.7323  
Iteration 2:   log likelihood = -2715.5419  
Iteration 3:   log likelihood = -2715.5387  
Iteration 4:   log likelihood = -2715.5387  

Truncated Poisson regression                      Number of obs   =       2125
Truncation point: 0                               LR chi2(16)     =     388.05
                                                  Prob > chi2     =     0.0000
Log likelihood = -2715.5387                       Pseudo R2       =     0.0667

---------------------------------------------------------------------------------------------
                 TNCHILDREN |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
----------------------------+----------------------------------------------------------------
             Marital_Status |
                         2  |  -.2737338   .0669437    -4.09   0.000     -.404941   -.1425265
                         3  |  -.0760975   .0550223    -1.38   0.167    -.1839392    .0317441
                            |
Femaleearningsshareofhouseh |  -.2387593   .0865392    -2.76   0.006     -.408373   -.0691456
           HOUSEHOLD_INCOME |  -.0311583   .0075559    -4.12   0.000    -.0459675   -.0163491
                        AGE |   .7477299   .2850515     2.62   0.009     .1890393     1.30642
                       age2 |  -.0185661   .0085606    -2.17   0.030    -.0353445   -.0017877
                       age3 |   .0001564   .0000843     1.86   0.063    -8.77e-06    .0003216
                 2.HISPANIC |  -.2643195   .1210433    -2.18   0.029    -.5015601    -.027079
   HISPANIC_BORN_OUT_THE_US |  -.1273851   .0968979    -1.31   0.189    -.3173014    .0625313
                1.White_NON |   .1437621   .1010381     1.42   0.155    -.0542689    .3417931
                  Asian_NON |          0  (omitted)
                1.Other_NON |   .3748508   .1382713     2.71   0.007      .103844    .6458575
                1.Black_NON |   .4289715    .112388     3.82   0.000      .208695    .6492479
                            |
                  EDUCATION |
                         2  |   .0537742     .06087     0.88   0.377    -.0655288    .1730772
                         3  |   .3164464   .0837348     3.78   0.000     .1523291    .4805636
                         4  |  -.0592162   .0774537    -0.76   0.445    -.2110227    .0925903
                         5  |   .0632101   .0579615     1.09   0.275    -.0503924    .1768125
                            |
                      _cons |  -9.010956   3.112718    -2.89   0.004    -15.11177   -2.910141
---------------------------------------------------------------------------------------------

          Comment

          • #6
            Does my result make sense?

            HTML Code:
            . lrtest nbreg poisson, force

Likelihood-ratio test                                 LR chi2(0)  =   -832.43
(Assumption: poisson nested in nbreg)                 Prob > chi2 =         .

. lrtest Poisson1 nbreg, force

Likelihood-ratio test                                 LR chi2(0)  =      0.00
(Assumption: nbreg nested in Poisson1)                Prob > chi2 =         .

            Comment

            • #7
              Dear Hm Saleh,

              I do not understand why you are comparing results from nbreg with results from truncated Poisson. In any case, using Poisson (not truncated) is likely to be preferable to NB, and the LR tests you report are unlikely to be valid.

              Best wishes,

              Joao
              • 1 like

              Comment

              Previous Next
              Working...
              X