[R] Change in 'solve' for r-patched

Roger Koenker roger at ysidro.econ.uiuc.edu
Sat Nov 1 19:48:46 CET 2003

For those possibly interested in reading something more about the rehabilitation of the
Cholesky decomposition following its long exile in the Siberia of numerical analysis
I would recommend Micheal Saunders brief:

\textsc{Saunders, M.~A.}  (1994): ``Major Cholesky would feel proud,''
  \emph{ORSA J. on Computing}, 6, 23--27.

url:	www.econ.uiuc.edu/~roger/my.html	Roger Koenker
email	rkoenker at uiuc.edu			Department of Economics
vox: 	217-333-4558				University of Illinois
fax:   	217-244-6678				Champaign, IL 61820

On 1 Nov 2003, Douglas Bates wrote:

> Thomas Lumley <tlumley at u.washington.edu> writes:
> chol2inv(crossprod(mm)) %*% crossprod(mm,y)
> is competitive in terms of timing, should be slightly more stable (it
> uses the Cholesky decomposition of the positive-definite matrix
> crossprod(mm) rather than an LU decomposition) and also produces the
> (X'X)^{-1} matrix.

More information about the R-help mailing list