Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: Matrix inversion "bug"

From   "FEIVESON, ALAN H. (AL) (JSC-SK) (NASA)" <>
To   "''" <>
Subject   RE: st: Matrix inversion "bug"
Date   Tue, 10 Jun 2003 15:45:29 -0500

Vince -

But on what basis can you say that A-1 is more "accurate" than [(A+A')/2 ]-1
(where A is the finite representation of the supposedly symmetric matrix
stored in Stata) when it is already known that A is supposed to be
symmetric? The very fact that you are using -symimv- in these 55 ado files
suggests that you are assuming A IS symmetric.

Al Feiveson

Vince wrote - 

"We occasionally use the trick that Mark mentions to symmetrize a matrix
inversion, but we never do it without first considering other options and
evaluating whether we believe it is computationally justified in the
context.  The function -syminv()- is used 127 times in 55 ado-files in
official Stata.  In all those uses, the matrix is pre-symmetrized in only 9
cases, and three of those are repeats for older version of code retained for
version control.  So official Stata has only 6 computations where
pre-symmetrizing the matrix was necessary and deemed safe."

*   For searches and help try:

© Copyright 1996–2019 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index