At 07:57 AM 2/14/2007, Maarten buis wrote:
2) I tend to trust the numerical properties of Stata functions more
then my own renderings of those functions.
Generally a good idea but not always. The ml routine for my oglm
program was having a terrible time converging when I used the
invcloglog function. I substituted the equivalent longer version of
the code and then it zoomed. I've found in other contexts that
seemingly innocuous changes in the code will occasionally make big
changes in how well the program functions. I often have no idea why
- maybe I've even uncovered some wildly esoteric bug - but I think it
is easier for me to make minor changes in my routines than to ask
Stata to rewrite theirs.
In the case of the code we were talking about earlier, not using
double precision was a clear and documented violation of recommended
programming practice (although unless you've read through Stata's ML
book, it doesn't surprise me that somebody would make that mistake).
-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
FAX: (574)288-4373
HOME: (574)289-5227
EMAIL: [email protected]
WWW (personal): http://www.nd.edu/~rwilliam
WWW (department): http://www.nd.edu/~soc
*
* 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/
