 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: logit results not matching with ml
I see the error now: You are giving the likelihood instead of the log
likelihood. Your program should be:
*------------ begin example -----------------
capture program drop mymle1
program define mymle1
version 8.0
args lnf theta1
quietly replace `lnf' = ln(exp(`theta1')/ /*
*/(1+exp(`theta1'))) /*
*/ if $ML_y1 == 1
quietly replace `lnf' = ln(1/(1+exp(`theta1'))) /*
*/ if $ML_y1 == 0
end
sysuse auto
ml model lf mymle1 (for = pr mpg wei)
ml search
ml maximize
logit for pri mpg wei
*------------- end example -------------------
Hope this helps,
Maarten
--- Maarten buis <[email protected]> wrote:
> I don't see any errors in your likelihood. You might want to change
> the
> scale of the variable pol_flow to measure whatever it measures in
> milions, i.e. devide pol_flow by 1000000. The reason is that the
> parameter for that variable is much much smaller than the others but
> the standard error is not dramatically large with respect to that
> estimate (though it is still insignigicant). This might cause
> numerical
> problems.
>
> Hope this helps,
> Maarten
>
> --- Tulika Narayan <[email protected]> wrote:
>
> > Hello,
> >
> > Stata version 8.0
> >
> > I am trying to run a non linear logit but before I do that, just to
> > test my
> > work I wrote the logit using stata's ml command but cannot get it
> to
> > replacate the logit results. In fact the likelihood does not even
> > converge.
> >
> > Here is the logit command I use
> >
> > logit green_present water_eda mixing_zone i_human temp_avg sal_avg
> > oxy_avg
> > coll_bottom oth_crab shell_present pol_flow
> >
> > The results
> >
> > Iteration 0: log likelihood = -37.362775
> > Iteration 1: log likelihood = -21.42294
> > Iteration 2: log likelihood = -19.458412
> > Iteration 3: log likelihood = -18.924528
> > Iteration 4: log likelihood = -18.641708
> > Iteration 5: log likelihood = -18.61557
> > Iteration 6: log likelihood = -18.615379
> >
> > Logit estimates Number of obs =
> > 58
> > LR chi2(10)
> =
> > 37.49
> > Prob > chi2
> =
> > 0.0000
> > Log likelihood = -18.615379 Pseudo R2 =
> > 0.5018
> >
> >
>
------------------------------------------------------------------------------
> > green_pres~t | Coef. Std. Err. z P>|z| [95%
> Conf.
> > Interval]
> >
>
-------------+----------------------------------------------------------------
> > water_edac~a | .0093596 .0085614 1.09 0.274 -.0074204
> > 0261396
> > mixing_zone | -.0115959 .0109941 -1.05 0.292 -.0331439
> > 009952
> > i_human | -.4338438 .4151365 -1.05 0.296 -1.247496
> > 3798089
> > temp_avg | -.5166932 .2147708 -2.41 0.016 -.9376362
> > -.0957502
> > sal_avg | -.0085234 .0685798 -0.12 0.901 -.1429375
> > 1258906
> > oxy_avg | -.4012752 .4969768 -0.81 0.419 -1.375332
> > 5727814
> > coll_bottom | -.1368686 .0534403 -2.56 0.010 -.2416096
> > -.0321276
> > oth_crab | -.8857099 .5262265 -1.68 0.092 -1.917095
> > 145675
> > shell_pres~t | .5593585 .4635706 1.21 0.228 -.3492232
> > 1.46794
> > pol_flow | 4.28e-07 4.20e-07 1.02 0.308 -3.95e-07
>
> > 1.25
> > e-06
> > _cons | 16.81088 7.657698 2.20 0.028 1.802072
> > 31.8197
> >
>
------------------------------------------------------------------------------
> >
> > note: 0 failures and 1 success completely determined.
> >
> > Here is the equivalent ml command.
> >
> > capture program drop mymle1
> > program define mymle1
> >
> > version 8.0
> > args lnf theta1
> > quietly replace `lnf' = exp(`theta1')/(1+exp(`theta1')) if
> > $ML_y1 ==
> > 1
> > quietly replace `lnf' = 1/(1+exp(`theta1')) if $ML_y1 == 0
> >
> > end
> >
> > local x1 "water_eda mixing_zone i_human temp_avg sal_avg oxy_avg
> > coll_bottom oth_crab shell_present pol_flow"
> > ml model lf mymle1 (green_present = `x1')
> > ml search
> >
> >
> > ml maximize, difficult
> >
> > Results
> >
> > initial: log likelihood = 29
> > alternative: log likelihood = 31.204268
> > rescale: log likelihood = 38
> > numerical derivatives are approximate
> > flat or discontinuous region encountered
> > Iteration 0: log likelihood = 38 (not concave)
> > Iteration 1: log likelihood = 40.018216 (not concave)
> > Iteration 2: log likelihood = 42.041076 (not concave)
> > Iteration 3: log likelihood = 43.042035 (not concave)
> > Iteration 4: log likelihood = 43.042049 (not concave)
> > Iteration 5: log likelihood = 43.042098 (not concave)
> > Iteration 6: log likelihood = 43.042103 (not concave)
> > Iteration 7: log likelihood = 43.042104 (not concave)
> >
> > Number of obs
> =
> > 58
> > Wald chi2(0)
> =
> > .
> > Log likelihood = 43.042104 Prob > chi2 =
> > .
> >
> >
>
------------------------------------------------------------------------------
> > green_pres~t | Coef. Std. Err. z P>|z| [95%
> Conf.
> > Interval]
> >
>
-------------+----------------------------------------------------------------
> > water_edac~a | .7304996 . . . .
> > .
> > mixing_zone | -45.97343 . . . .
> > .
> > i_human | .9961338 . . . .
> > .
> > temp_avg | 3.576235 . . . .
> > .
> > sal_avg | 10.37114 . . . .
> > .
> > oxy_avg | 2.229273 . . . .
> > .
> > coll_bottom | -.008225 . . . .
> > .
> > oth_crab | .0822529 . . . .
> > .
> > shell_pres~t | .3959996 . . . .
> > .
> > pol_flow | .0000723 . . . .
> > .
> > _cons | 64.28164 . . . .
> > .
> >
>
------------------------------------------------------------------------------
> >
> >
> > To help out the program, I gave the initial values from my logit
> >
> > ml init _cons =16.81 water_ = 0.009 mixing_zone = -0.115 i_human =
> > -0.4338
> > temp_avg =-0.516 sal_avg = -0.008 oxy_avg =-.401 coll_bott =-.136
> > oth_crab
> > =-.885 shell_p = 0.559 pol_flow = 0.00000428
> > ml maximize, difficult
> >
> > here are the results;
> >
> > Iteration 548: log likelihood = 50.644631 (backed up)
> > Iteration 549: log likelihood = 50.998166
> > Iteration 550: log likelihood = 50.999718
> > Iteration 551: log likelihood = 50.999727
> > Iteration 552: log likelihood = 50.999729 (backed up)
> >
> > Number of obs
> =
> > 58
> > Wald chi2(10)
> =
> > 0.03
> > Log likelihood = 50.999729 Prob > chi2 =
> > 1.0000
> >
> >
>
------------------------------------------------------------------------------
> > green_pres~t | Coef. Std. Err. z P>|z| [95%
> Conf.
> > Interval]
> >
>
=== message truncated ===
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
___________________________________________________________
Yahoo! Mail is the world's favourite email. Don't settle for less, sign up for
your free account today http://uk.rd.yahoo.com/evt=44106/*http://uk.docs.yahoo.com/mail/winter07.html
*
* 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/