Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: fweights + ml d0


From   "Alfonso Miranda Caso Luengo" <Alfonso.Miranda-Caso-Luengo@warwick.ac.uk>
To   <statalist@hsphsun2.harvard.edu>
Subject   Re: st: fweights + ml d0
Date   Mon, 14 Jun 2004 16:47:42 +0100

Dear all,

Here is an example of the problems I found with fw and ml d0. This code estimates a sequential probit. A dummy dependent variable is observed for each individual in various periods. The model predicts the probability of observing a given sequence. For each individual draws are iid. However, observations in the data set are not independent as each individual contributes a group of observations. Hence, the ml linear-form restriction is not met in this case.

Here is the example. Can anyone see the error?

Many thanks in advance,

Alfonso.


-------------------------------------------------------------------------------
       log:  C:\temp\mycode.smcl
  log type:  smcl
 opened on:  14 Jun 2004, 16:44:28

. do "C:\temp\mycode.do"

. 
. capture program drop mycode2

. program define mycode2, eclass
  1.         syntax [varlist] [if] [in] [fw] [,  I(string) ]
  2. 
.         /* Obtaining dependent variable and explanatory variables */    
. 
.         gettoken endgv exogv : varlist, parse("") 
  3. 
.         /* Selecting sample */  
. 
.         marksample touse
  4.         xt_iis `i'
  5.         local i "`s(ivar)'"
  6.         global S_ivar   "`i'"
  7.         markout `touse' `ivar' `varlist', strok
  8. 
.         /* Sorting data */
.         
.         sort `i'
  9. 
.         /* Etimate */
.         
.         ml model d0 mycode2_ll ("`endgv'": `endgv' = `exogv') /*
>         */ [`weight'`exp'] if `touse'
 10. 
.         ml check
 11. end

. 
. capture program drop mycode2_ll

. program define mycode2_ll
  1. args todo b lnf
  2. 
.  local y "$ML_y1"
  3.  local i "$S_ivar"
  4. 
.  tempvar xb F  last
  5. 
.  mleval `xb' = `b'
  6.  
.  qui { 
  7.         by `i' : gen `last' = cond(_n==_N,1,0)
  8. 
.         gen double `F' = . in 1  if  $ML_samp
  9.                        
.      #delimit ;
delimiter now ;
.      by `i': replace `F' = 
>      cond(_n==1,
>      cond(`y',normprob(`xb'),normprob(-`xb')),
>      cond(`y', 
>      normprob(`xb'),normprob(-`xb'))*`F'[_n-1] ) if $ML_samp;
 10.      #delimit cr
delimiter now cr
. 
.      mlsum `lnf' = log(`F')  if `last' 
 11. 
.    }
 12. end

. 
. mycode  depv d4-d30  catholic indspker edu12 c6372 c7382  agu-chih  durango-z
> a [fw=wt], i(id)

Test 1:  Calling mycode_ll to check if it computes log likelihood and
         does not alter coefficient vector...
         FAILED; mycode_ll returned error 499.

Here is a trace of its execution:
------------------------------------------------------------------------------
-> mycode_ll 0 ML_b ML_f __0000QH __0000QI __0000QA
- `begin'
- args todo b lnf
- local y "$ML_y1"
- local i "$S_ivar"
- tempvar xb F last
- mleval `xb' = `b'
- qui {
- by `i' : gen `last' = cond(_n==_N,1,0)
- gen double `F' = . in 1 if $ML_samp
- by `i': replace `F' = cond(_n==1, cond(`y',normprob(`xb'),normprob(-`xb')), c
> ond(`y', normprob(`xb'),normprob(-`xb'))*`F'[_n-1] ) if $ML_samp
- mlsum `lnf' = log(`F') if `last'
mlsum must be over entire estimation sample when weights
are specified
  }
- `end'
------------------------------------------------------------------------------
Fix mycode_ll.
r(499);

end of do-file
r(499);

. log close
       log:  C:\temp\mycode.smcl
  log type:  smcl
 closed on:  14 Jun 2004, 16:46:01
-------------------------------------------------------------------------------

=====================================
Alfonso Miranda
PhD Student

Economics Department
University of Warwick
Coventry CV4 7AL
E-mail:ecrgw@warwick.ac.uk
WebPage: http://www.warwick.ac.uk/go/amiranda
=====================================

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index