Re: st: logistic reg version 8 & 11

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Fri, 10/9/10, Kristin MacDonald, StataCorp wrote:
> In Stata 11, the -logistic- command is evaluating the same
> likelihood as in previous versions.  The only difference is
> the path it may take on its way to the solution.  Prior to
> Stata 11, -logit- (and hence -logistic-) used Stata's
> internal optimizer, which does not employ the telescoped
> stepping implemented in -ml-.  In Stata 11, -logit- now
> uses -ml- to fit logistic regression (but preserves the
> original algorithm under version control).

In tough estimation problem, like the one below, -logit-  
seems to converge to slightly different solutions. The 
difference is small enough that it does not worry me, but 
large enough to be picked up by the default setting in 
-mkassert-.

*------------- begin example ----------------
sysuse nlsw88, clear

recode occupation (11/12=4) ///
                  (9/10=13)
				  
tab occupation, gen(occ)
drop occ1 // reference category is occ1


version 11.1
logit union occ? married never_married south
matrix b10 = e(b)
matrix v10 = e(V)
version 10
logit union occ? married never_married south
matrix b11 = e(b)
matrix v11 = e(V)

// difference larger than 1E-8 (the default in -mkassert-)
di mreldif(b10,b11)
di mreldif(v10,v11) 

// the problem is occ7 
matrix dif = b10 - b11
matrix list dif

// this should be a difficult category
// to get an estimate of
tab occ7 union
*------------------- end example ---------------------

-- Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   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/