Statalist The Stata Listserver

[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: possible bug in the singular value decomposition (matrix svd)

From   "Sam Thompson" <>
To   <>
Subject   st: possible bug in the singular value decomposition (matrix svd)
Date   Thu, 31 Aug 2006 10:07:49 -0400

Hi everybody. I'm running constrained regressions with nearly singular
covariate matricies. I use the singular value decomposition to produce
generalized inverses which are numerically stable. I think that I have
found a problem with the matrix svd command in STATA 8. The command
syntax is

matrix svd U M V = A

The resulting matricies should satisfy the equality A = U*diag(M)*V'.
For a particular matrix, I find that the equality fails beyond the usual
errors due to numerical imprecision. When I feed exactly the same matrix
into the mata "fullsvd" command in STATA 9, the problem is resolved.
However, for a number of reasons, we need to keep using STATA 8 here.

So: is there a known bug in the matrix svd command? If so, is there a
known fix?

Thanks a lot,

Sam Thompson, Arrowstreet Capital, LLP

If anyone wants, I can send the matrix along. It's not too big - 62 x 60

This message is intended solely for the designated recipient(s). It may contain confidential or proprietary information and may be subject to confidentiality protections. If you are not a designated recipient, you may not review, copy, or distribute this message. If you receive this in error, please notify the sender by reply e-mail and delete this message. 

Arrowstreet Capital, L.P. is an Equal Opportunity Employer. 

*   For searches and help try:

© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index