st: Mata, xtabond2, memory, matsize, and outreg2

["roy wada" <roywada@hotmail.com>]

It was recently brought up by R. Burt Porter that outreg2 could not always 
read e(b) after xtabond2. Burt kindly sent me his do-file with the data. I 
found that under certain circumstances xtabond2 produce e(b) and e(V) 
matrices that cannot be accessed by normal means ("op. sys. refuses to 
provide memory"). The problems will go away if Mata is turned off with 
-nomata- option.

It appears recent changes have been made to the executable concerning Mata 
and memory allocation, possibly with xtabond2 in mind, inlcuding an 
option-like-somthing called xtabond2_mata().

The latest executable (downloaded today) also appears to be overly 
sensitivie to the memory and matsize allocation. It seems to crash with the 
message, "memroy allocation failed," when the memory and matsize are 
repeatedly changed in search of the maximum allowable limit. A slightly 
lesser maximum memory appears to be available with the latest executable. 
(Where does it go?)

Exactly which part of Mata, xtabond2, memory, and matsize is causing the 
problem I have no clue.

(A note to the author of xtabond: the date checking at the top of the 
program no longer works; it will crash for a slightly older executable from 
April 2006.)

Roy



> Thank you for your response, but my problem is not with loading a large 
> dataset.  I ran a trace on outreg2 and found the memory error occurs in a 
> matrix assignment of the coefficient vector e(b).  Using the command 
> "matrix list e(b)" from the command line will also return the error.  e(b) 
> is relatively a relatively small 47x1 vector.
*
*   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/