Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: Maximum likelihood with loops


From   Olga Lebedeva <lebedeva@corporate-finance-mannheim.de>
To   statalist@hsphsun2.harvard.edu
Subject   st: Maximum likelihood with loops
Date   Fri, 26 Jun 2009 18:12:48 +0200

Dear Stata users,

I am running into a problem when I need to conduct maximum likelihood 
estimates
on subsamples of the data. I use "if" option in ml model and run loops 
over different
subsamples.

My problem is as follows.
The model converges in general when I run it on different subsamples one 
by one
(entered manually). However, in order to automatize the procedure, I use 
loops -
and get estimates only for the first subsample of data. when the program 
moves on to the second
subsample, error message 430 "missing values encountered" is displayed. 
However,
the model converges if I start with second subsample but will not 
converge for the third, fourth, fifth etc.

does anyone know what the reson for the problem might be?

ML code:

program pin_lf_02

version 1.2

args lnf x y l k
tempvar mu eps alpha delta P1 P2 P3 P4
quietly {

        gen double `eps'=exp(`x')  
        gen double `mu'=exp(`y')
        gen double `alpha'=1/(1+exp(-`l'))
        gen double `delta'=1/(1+exp(-`k'))
        gen double `P1'= - `eps' + $ML_y1*ln(`eps') - lngamma($ML_y1 + 1)
        gen double `P2'= - `eps' + $ML_y2*ln(`eps') - lngamma($ML_y2 + 1)
        gen double `P3'= -(`mu'+`eps') + $ML_y2*ln(`mu'+`eps') - 
lngamma($ML_y2 + 1)
        gen double `P4'=-(`mu'+`eps')+ $ML_y1*ln(`mu'+`eps') - 
lngamma($ML_y1 + 1)
       
        replace 
`lnf'=ln(exp(ln(1-`alpha')+`P1'+`P2')+exp(ln(`alpha')+ln(`delta')+`P1'+`P3')+exp(ln(`alpha')+ln(1-`delta')+`P4'+`P2'))



}

end

forvalues q = 1/20 {

forvalues p = 528/537 {



ml model lf pin_lf_02 (x: buy sell = ) (y: buy sell = ) (l:  buy sell = 
) (k: buy sell = ) if (((monthyear==`p') | (monthyear==`p'+1) | 
(monthyear==`p'+2))&(num1==`q')), missing technique(nr bhhh)
ml init /x = 3 /y = 3 /l = 0 /k = 0
* ml check
ml search x -1 6 y -1 5 l -3 3 k -3 3
* ml search

ml maximize, difficult

mat beta`p' = e(b)
    replace x1 = el(beta`p',1,1) if (monthyear==`p'+2) & (num1==`q')
    replace y1 = el(beta`p',1,2) if monthyear==`p'+2 & (num1==`q')
    replace l1 = el(beta`p',1,3) if monthyear==`p'+2 & (num1==`q')
    replace k1 = el(beta`p',1,4) if monthyear==`p'+2 & (num1==`q')
}

}

-- 
________________________________
<br>
Olga Lebedeva 
<br>
PhD - Candidate in Finance <br>
Universität Mannheim <br>
L9, 1-2 (Zimmer 404), 68131 Mannheim <br>
Tel: 0621-181-2278



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   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