I'm interested in the answer to this too. In the astrodynamics world at least this is called sequential-batch least squares and involves saving the information matrix (inverse of the covariance) and the transformed residuals. However, you might want to look into Kalman filters which do this and other things (not in GSL as far as I know).
On Mon, Apr 7, 2008 at 2:15 PM, António Alegria <[EMAIL PROTECTED]> wrote: > Hi, > > I'm considering using the GSL for least squares fitting to represent a > continuously updated time series. Instead of saving each new (x,f(x)) and > running the regression over all collected data, I would like to > incrementally update the model with the most recent reading. Is this > possible with GSL? > > Thanks! > _______________________________________________ > Help-gsl mailing list > [email protected] > http://lists.gnu.org/mailman/listinfo/help-gsl > _______________________________________________ Help-gsl mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-gsl
