Hi Antonio,
Your estimation of the variance has some errors. If you make
it like this
gen v= ((ntcrimes/pop)*((pop-ntcrimes)/pop))/(pop*(normden(lpe1)^2))
you should get the appropriate results..
You can see how i reproduced the results below
clear
set obs 500000
generate x1 = abs(invnorm(uniform()))
generate x2 = abs(invnorm(uniform()))
generate lp = 1 + .5*x1 - .25*x2
generate mu = exp(lp)
rndpoix mu
rename xp ntcrimes
gen pop = round(100*uniform())+ ntcrimes
gen double p=ntcrimes/pop
gen double lpe1 =invnorm(p)
gen v= ((ntcrimes/pop)*((pop-ntcrimes)/pop))/(pop*(normden(lpe1)^2))
gprobit ntcrimes pop x1 x2
regress lpe1 x1 x2 [aweight=1/v]
gprobit ntcrimes pop x1 x2
Weighted least-squares probit estimates for grouped data
Source | SS df MS Number of obs =
472531
-------------+------------------------------ F( 2,472528)
=29169.51
Model | 10652.8642 2 5326.4321 Prob > F =
0.0000
Residual | 86284.8961472528 .182602716 R-squared =
0.1099
-------------+------------------------------ Adj R-squared =
0.1099
Total | 96937.7603472530 .205146256 Root MSE =
.42732
----------------------------------------------------------------------------
--
ntcrimes | Coef. Std. Err. t P>|t| [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
x1 | .2110069 .0009378 224.99 0.000 .2091688
.212845
x2 | -.0964621 .0010868 -88.76
0.000 -.0985921 -.0943321
_cons | -1.546212 .0013383 -1155.35
0.000 -1.548835 -1.543589
----------------------------------------------------------------------------
--
. regress lpe1 x1 x2 [aweight=1/v]
(sum of wgt is 6.1780e+06)
Source | SS df MS Number of obs =
472531
-------------+------------------------------ F( 2,472528)
=29169.51
Model | 10652.8642 2 5326.43212 Prob > F =
0.0000
Residual | 86284.8962472528 .182602716 R-squared =
0.1099
-------------+------------------------------ Adj R-squared =
0.1099
Total | 96937.7604472530 .205146256 Root MSE =
.42732
----------------------------------------------------------------------------
--
lpe1 | Coef. Std. Err. t P>|t| [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
x1 | .2110069 .0009378 224.99 0.000 .2091688
.212845
x2 | -.0964621 .0010868 -88.76
0.000 -.0985921 -.0943321
_cons | -1.546212 .0013383 -1155.35
0.000 -1.548835 -1.543589
----------------------------------------------------------------------------
--
Regards
Anthony
----- Original Message -----
From: "Antonio Rodrigues Andres" <[email protected]>
To: <[email protected]>
Sent: Friday, October 24, 2003 9:43 AM
Subject: st: Programming question
> Dear friends
>
>
> I run this regression using probit grouped data regression models
>
> gprobit ntcrimes pop cl income1 pyoung pforeign unempl $zvars
>
> I have tried to replicate the analysis
>
> gen lpe1=invnorm(ntcrimes/pop)
> gen v= (ntcrimes/pop)*(pop-ntcrimes/pop)/(pop*normden^2*lpe1)
>
> regress lpe1 cl income1 pyoung pforeign unempl $zvars [aweight=1/v]
>
>
> but I dont get the same results I guess the problem is concerning with
> the weights
>
> Antonio
> *
> * 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/
