st: Re: problems in maximizing a likelihood fn

[Kit Baum <baum@bc.edu>]

Although I do not doubt that Deepankar stumbled onto some alteration  
of his program that made it work, the conclusion drawn below is  
absolutely invalid. Stata scalars have EXACTLY the same precision as  
variables, temporary or permanent, of data type double. Try

clear
set obs 100
scalar p = _pi
gen double pp = _pi
gen double diff = pp - p
summ pp diff
di %20.16f p
format pp %20.16f
l pp in 1

and you will see that the two have identical capabilities in terms of  
precision. So the advice below -- which could be moderately  
disastrous to a user with a large N, counseled to create 10 or 20  
variables of length N to hold 10 or 20 constant values in her ML  
evaluator -- should be disregarded.

Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html


On Jun 16, 2006, at 2:33 AM, Deepankar wrote:


> Moral of the story: never use scalars in -ml- estimation; use  
> temporary
> variables (if you need them) and generate them as -double-. And  
> only as
> - -double-.
*
*   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/