R help - Feb 2010 - iterative regressions, adding a new line of data each, time

Possibly of limited use to the original poster, but of interest more generally,
there are
a number of tools developed in the 70s for updating the matrix decompositions.
There's at
least one in my 1979 book "Compact numerical methods for computers"
(still in print
apparently) -- we didn't have enough space to keep all the data in the
HP9830A machine. If
there's sufficient interest, I can probably come up with some suitable
references.

Beyond the historical interest, however, it may be that the updating of the
decompositions
is useful in improving efficiency of computation and revealing structure of the
problems.
The cost, of course, is in writing the R interface code, so unless one needs to
solve a
lot of problems, I'd use R's resources, though possibly via svd.

JN

On Tue, 2 Feb 2010, Prof. John C Nash wrote:
> Possibly of limited use to the original poster, but of interest more 
> generally, there are a number of tools developed in the 70s for updating
the
> matrix decompositions. There's at least one in my 1979 book
"Compact
> numerical methods for computers" (still in print apparently) -- we
didn't
> have enough space to keep all the data in the HP9830A machine. If
there's
> sufficient interest, I can probably come up with some suitable references.
The biglm package does precisely this.   It uses an incremental QR
decomposition, based on Fortran code by Alan Miller.

       -thomas


Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle