One answer is: by very clever C code that is proprietary to StataCorp.
Nick
[email protected]
Jacob Felson
I understand that computing with very large matrices is inefficient, and
often impossible, given memory constraints. That said, I am very
curious about how commands such as matrix accum carry out matrix
algebra. I wanted to program an iterated least squares algorithm to
estimate a mixed model, but I found that I could not do so (either in
stata or mata) by brute-force matrix computations. The matrices were
simply too large. So how do Stata and other programs go about matrix
computation in a way that does not swamp memory?
*
* 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/
