And one more semi-related thing:
When I use the endogenous variable in it's continuous form (and a
square form) I get endless iterations until I give up. See below:
. ivprobit employment (centbmi centbmisq=diet1 diet2) centage
centagesq edu2 edu3 edu4 edu5 edu6 marriag
> e2 marriage3 marriage4 children
Fitting exogenous probit model
Iteration 0: log likelihood = -597.05914
Iteration 1: log likelihood = -415.52928
Iteration 2: log likelihood = -402.8589
Iteration 3: log likelihood = -402.40821
Iteration 4: log likelihood = -402.40667
Iteration 5: log likelihood = -402.40667
Fitting full model
Iteration 0: log likelihood = -9748.1502 (not concave)
Iteration 1: log likelihood = -9582.4934 (not concave)
Iteration 2: log likelihood = -9553.7582 (not concave)
Iteration 3: log likelihood = -9547.7161 (not concave)
Iteration 4: log likelihood = -9537.2101 (not concave)
Iteration 5: log likelihood = -9526.3101 (not concave)
Iteration 6: log likelihood = -9525.619 (not concave)
Iteration 7: log likelihood = -9524.4498 (not concave)
Iteration 8: log likelihood = -9522.9543 (not concave)
Iteration 9: log likelihood = -9518.4009 (not concave)
Iteration 10: log likelihood = -9517.5952 (not concave)
Iteration 11: log likelihood = -9516.7487 (not concave)
Iteration 12: log likelihood = -9512.821 (not concave)
Iteration 13: log likelihood = -9512.2921 (not concave)
Iteration 14: log likelihood = -9512.1517 (not concave)
Iteration 15: log likelihood = -9511.7616 (not concave)
Iteration 16: log likelihood = -9500.6588 (not concave)
Iteration 17: log likelihood = -9499.9541 (not concave)
Iteration 18: log likelihood = -9477.155 (not concave)
Iteration 19: log likelihood = -9475.7283 (not concave)
Iteration 20: log likelihood = -9475.0241 (not concave)
Iteration 21: log likelihood = -9396.2923 (not concave)
Iteration 22: log likelihood = -9396.2152 (not concave)
Iteration 23: log likelihood = -9393.6135 (not concave)
Iteration 24: log likelihood = -9393.4571 (not concave)
Iteration 25: log likelihood = -9383.8382 (not concave)
Iteration 26: log likelihood = -9382.7582 (not concave)
Iteration 27: log likelihood = -9380.8916 (not concave)
Iteration 28: log likelihood = -9380.044 (not concave)
Iteration 29: log likelihood = -9378.4173 (not concave)
Iteration 30: log likelihood = -9376.8372 (not concave)
Iteration 31: log likelihood = -9376.0801 (not concave)
Iteration 32: log likelihood = -9375.4403 (not concave)
Iteration 33: log likelihood = -9374.4594 (not concave)
Iteration 34: log likelihood = -9374.4547 (not concave)
Iteration 35: log likelihood = -9372.391 (not concave)
Iteration 36: log likelihood = -9371.909 (not concave)
Iteration 37: log likelihood = -9364.3699 (not concave)
Iteration 38: log likelihood = -9363.6614 (not concave)
Iteration 39: log likelihood = -9363.2954 (not concave)
Iteration 40: log likelihood = -9363.2702 (not concave)
Iteration 41: log likelihood = -9362.4889 (not concave)
Iteration 42: log likelihood = -9357.4852 (not concave)
Iteration 43: log likelihood = -9356.6759 (not concave)
Iteration 44: log likelihood = -9353.7967 (not concave)
Iteration 45: log likelihood = -9353.6053 (not concave)
Iteration 46: log likelihood = -9353.4411 (not concave)
Iteration 47: log likelihood = -9350.254 (not concave)
Iteration 48: log likelihood = -9342.8995 (not concave)
Iteration 49: log likelihood = -9341.0832 (not concave)
Iteration 50: log likelihood = -9340.873 (not concave)
Iteration 51: log likelihood = -9329.6951 (not concave)
Iteration 52: log likelihood = -9329.658 (not concave)
Iteration 53: log likelihood = -9329.1022 (not concave)
Iteration 54: log likelihood = -9328.9638 (not concave)
Iteration 55: log likelihood = -9328.8256 (not concave)
Iteration 56: log likelihood = -9328.0734
Iteration 57: log likelihood = -9327.8131 (not concave)
Iteration 58: log likelihood = -9327.6051 (not concave)
Iteration 59: log likelihood = -9327.5932
Iteration 60: log likelihood = -9327.3608 (not concave)
Iteration 61: log likelihood = -9327.3484 (not concave)
Iteration 62: log likelihood = -9327.3454 (not concave)
Iteration 63: log likelihood = -9327.3438 (not concave)
Iteration 64: log likelihood = -9327.3425 (not concave)
Iteration 65: log likelihood = -9327.3417
Iteration 66: log likelihood = -9327.3415 (not concave)
Iteration 67: log likelihood = -9327.34 (not concave)
Iteration 68: log likelihood = -9327.3396 (not concave)
Iteration 69: log likelihood = -9327.3393
Iteration 70: log likelihood = -9327.3378 (not concave)
Iteration 71: log likelihood = -9327.3378 (not concave)
Iteration 72: log likelihood = -9327.3378 (not concave)
Iteration 73: log likelihood = -9327.3378
Iteration 74: log likelihood = -9327.3375 (not concave)
Iteration 75: log likelihood = -9327.3374 (not concave)
Iteration 76: log likelihood = -9327.3374
Iteration 77: log likelihood = -9327.337
Iteration 78: log likelihood = -9327.3368 (not concave)
Iteration 79: log likelihood = -9327.3364
Iteration 80: log likelihood = -9327.3359 (not concave)
Iteration 81: log likelihood = -9327.3359
Iteration 82: log likelihood = -9327.3355 (not concave)
Iteration 83: log likelihood = -9327.3354
Iteration 84: log likelihood = -9327.3351
Iteration 85: log likelihood = -9327.3348 (not concave)
Iteration 86: log likelihood = -9327.3348 (not concave)
Iteration 87: log likelihood = -9327.3347 (not concave)
Iteration 88: log likelihood = -9327.3347
Iteration 89: log likelihood = -9327.3347
Iteration 90: log likelihood = -9327.3343 (not concave)
Iteration 91: log likelihood = -9327.3343 (not concave)
Iteration 92: log likelihood = -9327.3343
Iteration 93: log likelihood = -9327.3341 (not concave)
Iteration 94: log likelihood = -9327.3341 (not concave)
Iteration 95: log likelihood = -9327.334
--Break--
r(1);
On 9/13/05, Tinna <[email protected]> wrote:
> Hmmmm!
>
> I appreciate your time guys.
>
> Here are my commands and results.
>
> . ivprobit employment ( OBESEbmi=diet1 diet2) centage centagesq
> edu2 edu3 edu4 edu5 edu6 marriage2 marri
> > age3 marriage4 children
>
>
> Fitting exogenous probit model
>
> Iteration 0: log likelihood = -597.05914
> Iteration 1: log likelihood = -415.87674
> Iteration 2: log likelihood = -403.33373
> Iteration 3: log likelihood = -402.89898
> Iteration 4: log likelihood = -402.89762
> Iteration 5: log likelihood = -402.89762
>
> Fitting full model
>
> Iteration 0: log likelihood = -746.36499
> Iteration 1: log likelihood = -744.94761
> Iteration 2: log likelihood = -744.38178
> Iteration 3: log likelihood = -744.38
> Iteration 4: log likelihood = -744.38
>
> Probit model with endogenous regressors Number of obs = 1045
> Wald chi2(12) = 281.90
> Log likelihood = -744.38 Prob > chi2 = 0.0000
>
> ------------------------------------------------------------------------------
> | Coef. Std. Err. z P>|z| [95% Conf. Interval]
> -------------+----------------------------------------------------------------
> employment |
> OBESEbmi | -.8835098 .4013275 -2.20 0.028 -1.670097 -.0969223
> centage | -.0114935 .0039625 -2.90 0.004 -.0192598 -.0037272
> centagesq | -.0026056 .0002318 -11.24 0.000 -.0030599 -.0021513
> edu2 | -.5586423 .1567106 -3.56 0.000 -.8657895 -.2514951
> edu3 | .152094 .1354019 1.12 0.261 -.1132888 .4174769
> edu4 | .4574337 .1725559 2.65 0.008 .1192304 .795637
> edu5 | .4399531 .1929985 2.28 0.023 .0616829 .8182233
> edu6 | .9912227 .3626966 2.73 0.006 .2803504 1.702095
> marriage2 | -.0014036 .181648 -0.01 0.994 -.3574272 .3546199
> marriage3 | -.4220406 .1645533 -2.56 0.010 -.7445591 -.099522
> marriage4 | -.6121703 .2256473 -2.71 0.007 -1.054431 -.1699096
> children | -.0192385 .0425978 -0.45 0.652 -.1027286 .0642516
> _cons | 1.534364 .1681024 9.13 0.000 1.20489 1.863839
> -------------+----------------------------------------------------------------
> /lnsigma | -1.090328 .0218741 -49.85 0.000 -1.133201 -1.047456
> /athrho | .3088546 .1551784 1.99 0.047 .0047107 .6129986
> -------------+----------------------------------------------------------------
> sigma | .3361062 .007352 .322001 .3508292
> rho | .2993948 .1412686 .0047106 .5462345
> ------------------------------------------------------------------------------
> Instrumented: OBESEbmi
> Instruments: centage centagesq edu2 edu3 edu4 edu5 edu6 marriage2 marriage3
> marriage4 children diet1 diet2
> ------------------------------------------------------------------------------
> Wald test of exogeneity (/athrho = 0): chi2(1) = 3.96 Prob > chi2 = 0.0466
>
> . mfx
>
> warning: predict() expression unsuitable for standard-error calculation;
> option nose imposed
>
>
> Marginal effects after ivprobit
> y = Fitted values (predict)
> = .77898262
> -------------------------------------------------------------------------------
> variable | dy/dx X
> ---------------------------------+---------------------------------------------
> OBESEbmi | -.8835098 .159183
> centage | -.0114935 -.21666
> centagesq | -.0026056 251.145
> edu2 | -.5586423 .125493
> edu3 | .152094 .228086
> edu4 | .4574337 .159307
> edu5 | .4399531 .144449
> edu6 | .9912227 .056753
> marriage2 | -.0014036 .099244
> marriage3 | -.4220406 .085597
> marriage4 | -.6121703 .061284
> children | -.0192385 2.40813
> diet1 | 0 .613461
> diet2 | 0 .150254
> -------------------------------------------------------------------------------
>
>
> On 9/13/05, Scott Merryman <[email protected]> wrote:
> >
> > > -----Original Message-----
> > > From: [email protected] [mailto:owner-
> > > [email protected]] On Behalf Of Tinna
> > > Sent: Tuesday, September 13, 2005 3:37 PM
> > > To: [email protected]
> > > Subject: Re: st: RE: Marginal effects after ivprobit
> > >
> > > Thanks for the answer Scott. Yes I am pretty sure.
> > >
> > > If you try the same estimations again without regressing quietly then
> > > you will probably see that the coefficients you get after mfx are the
> > > same as from the original estimation.
> >
> > <snip>
> >
> > But they are not.
> >
> > . webuse laborsup, clear
> >
> > . ivprobit fem_work fem_educ kids (other_inc = male_educ) , nolog
> >
> > Probit model with endogenous regressors Number of obs = 500
> > Wald chi2(3) = 163.88
> > Log likelihood = -2368.2062 Prob > chi2 = 0.0000
> >
> > ----------------------------------------------------------------------------
> > Coef. Std. Err. z P>|z| [95% Conf. Interval]
> > -------------+--------------------------------------------------------------
> > fem_work |
> > other_inc | -.0542756 .0060854 -8.92 0.000 -.0662027 -.04234
> > fem_educ | .211111 .0268648 7.86 0.000 .1584569 .26376
> > kids | -.1820929 .0478267 -3.81 0.000 -.2758316 -.08835
> > _cons | .3672083 .4480724 0.82 0.412 -.5109975 1.2454
> > -------------+--------------------------------------------------------------
> > /lnsigma | 2.813383 .0316228 88.97 0.000 2.751404 2.8753
> > /athrho | .3907858 .1509443 2.59 0.010 .0949403 .68663
> > -------------+--------------------------------------------------------------
> > sigma | 16.66621 .5270318 15.66461 17.731
> > rho | .3720374 .1300519 .0946561 .59581
> > ----------------------------------------------------------------------------
> > Instrumented: other_inc
> > Instruments: fem_educ kids male_educ
> > ----------------------------------------------------------------------------
> > Wald test of exogeneity (/athrho = 0): chi2(1) = 6.70 Prob > chi2 =
> > 0.0096
> >
> > . estimates store iv
> >
> > . mfx, predict(p)
> >
> > warning: predict() expression p unsuitable for standard-error calculation;
> > option nose imposed
> >
> >
> > Marginal effects after ivprobit
> > y = Probability of positive outcome (predict, p)
> > = .44363395
> > ----------------------------------------------------------------------------
> > variable | dy/dx X
> > ---------------------------------+------------------------------------------
> > other_inc | -.0214364 49.6023
> > fem_educ | .0833791 12.046
> > kids | -.0719183 1.976
> > male_educ | 0 11.966
> > ----------------------------------------------------------------------------
> >
> > . estimates store mfx
> >
> > Or, all together for easy comparison:
> >
> > . estout iv mfx, style(fixed) margin meqs(fem_work) label varwidth(24)
> > varlabels(_cons Constant) keep(fem_work:) collabels(,none)
> >
> > iv mfx
> > Does female work?
> > Other income -.0542756 -.0214364
> > Female education level .211111 .0833791
> > Number of children -.1820929 -.0719183
> > Constant
> >
> >
> > Scott
> >
> >
> >
> >
> > *
> > * For searches and help try:
> > * http://www.stata.com/support/faqs/res/findit.html
> > * http://www.stata.com/support/statalist/faq
> > * http://www.ats.ucla.edu/stat/stata/
> >
>
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
