On 3 March 2016 at 02:07, Lubomir I. Ivanov <[email protected]> wrote: > > so for instance, the tables start at 7 bars and end at 276 bars, but > if the input is 1 bar or something like 500 bars, the value might not > follow the fitted curve to the data set and deviate. this is usually > worse with something like Lagrangian interpolation, tolerable to quite > bad with polynomial regression and pretty good with splines (but > splines are expensive and i don't have a ready algorithm / website for > them). >
erm, bad wording. Lagrangian interpolation is a polynomial regression. i meant, compared to something like the least-squares method. the Lagrange polynomial tends to deviate *a lot*. lubomir -- _______________________________________________ subsurface mailing list [email protected] http://lists.subsurface-divelog.org/cgi-bin/mailman/listinfo/subsurface
