[SciPy-User] weighted griddata

Benjamin Root ben.root@ou....
Fri Sep 17 15:03:23 CDT 2010


On Fri, Sep 17, 2010 at 2:00 PM, Christopher Barker
<Chris.Barker@noaa.gov>wrote:

> Adam Ryan wrote:
> > For clarification, all the points of a line have the same t value, but
> > each point has a weight of 0 to 1.
>
>
> does that mean the ones with weight 0 you want to ignore, and ones with
> weight 1 you want to fit exactly? i.e. you know which ones are more
> robust/accurate?
>
> If so then it doesn't sound to me like any of the smoothing routines
> being talked about are going to do the right thing, at least not out of
> the box -- I think they all assume that all points are equally valid,
> and weight according to how far away points are, or, more generally, how
> well they fit a smooth function.
>
> Of the top of me head, I imagine you may be able to do some sort of
> least squares type fit to a known function (maybe a polynomial or
> piecewise-polynomial), but with the error term weighted according to to
> your known weights. That could look a lot like a spline fit, but you'd
> have to insert your weighting in there somehow.
>
> -Chris
>
>
>
>
Come to think of it, wouldn't something like 3DVar or 4DVar (data
assimilation) be more what he is looking for?  I know in meteorology, we use
various data assimilation techniques to not only "snap" the observation data
to a regular grid, but to also give proper weight to data that is more
reliable than others.

Ben Root
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.scipy.org/pipermail/scipy-user/attachments/20100917/ae16e131/attachment.html 


More information about the SciPy-User mailing list