Re: Missing Values

Hi Cameron,

> We dont quite understand:
> > Then you can resample that IrregularSet back to your original GriddedSet.
> 
> This is what we have done:
> 1. create new griddedset of points have non-missing values.

This is fine if the points with non-missing values form a
gridded topology.

> 2. create new field from original by resampling with new set.
> 3. create new irregular set of points we want.
> 4. create new field by resampling with new irregular set.
> 
> This seems to work quite well for interpolation but does not fill in values
> for points on the end of the irregular set that have missing values
> (i.e. extrapolate). I only need to fill in the end samples with
> nearest neighbor or something simmilar. Can this be done?

Good point.  Field.resample() interpolates bit does not extrapolate
using the current Sets (their valueToInterp() methods define the
interpolation algorithm).  You can either extrapolate by writing
custom code, or by extending the Set class that you're using for
your non-missing values, and over-riding valueToInterp to
extrapolate.

> Your mail seemed to suggest that we use an irregular set and then a gridded
> set however this gives a nullpointer exception at
> irregular1dset.valueToInterp();

This is a bug.  I'd be very interested in getting a stack dump of
this with line numbers (running under jdb).  I may also need to
run your program here to find this bug.

Cheers,
Bill
----------------------------------------------------------
Bill Hibbard, SSEC, 1225 W. Dayton St., Madison, WI  53706
hibbard@xxxxxxxxxxxxxxxxx  608-263-4427  fax: 608-263-6738
http://www.ssec.wisc.edu/~billh/vis.html