Boyce Work

Beth Whittle's powerpoint summarizing our work so far....
The Boyce challenge results: Here are All reconstructions!
```
Here's my choice of species to analyze--I won't tell you why...

Carex elynoides (#15)
Carex scopularum (#17)
Lewisia pygmaea (#40)
Pentaphylloides floribunda (#58)
Saxifraga rhomboidea (#62)
```
Here's how I'm producing these plots: I use boyce_estimate_all.lsp, and the results of kriging each spatially autocorrelated factor
along with the means of all the other variables.
This file writes two files: "species.R" and "rawdata.R". The first contains the map of the species, the second the raw data for that species (the data locations). Actually, I've now written out each file with its name Rboyce/species.n.R, etc., where n is the species number. E.g., species.1.R.
Those files are then read by the R script BethMass.R, which creates the final image.
Incorporating the anisotropy: variograms and kriged results.
- kriged map of species 2 interpolated using anisotropic variogram (Here reconstructed using interpolated factors)
- kriged map of species 9 interpolated using anisotropic variogram (Here reconstructed using interpolated factors)
- kriged map of species 45 interpolated using anisotropic variogram (Here reconstructed using interpolated factors)
- kriged map of species 29 interpolated using anisotropic variogram (Here reconstructed using interpolated factors)
- Kernel-smoothed Elevation estimates (used as the grid for 3d-kriging)
  The following variogram plots are mislabelled. Sorry about that!
- Kernel estimation of Factor 1 (variogram)
  - Looks like it might require a periodic model (local copy)
  - Kernel estimation of Factor 1 using cosine (variogram)
- Kernel estimation of Factor 2 (variogram)
- Kernel estimation of Factor 3 (variogram)
- Factor 4 didn't seem to show any spatial autocorrelation
- Kernel estimation of Factor 5 (variogram)
- Factor 6 didn't seem to show any spatial autocorrelation
- Factor 7 nada
- Factor 8 nada
- Factor 9 nada
Here is the set of 3-d (isotropic) variograms, obtained incorporating elevation. Many are worth looking into.
Beth's page of 2d stuff (dead link -- lost? Or can I get that back?)
Original data
using xlispstat

I'd made a mistake in the variograms I was showing before, Beth, so make a note of that -- you weren't crazy! I was....
Here's the transformation via Euler angles (local copy) that I've carried out. This site gives them an intuitive meaning, which is interesting. I think that they're got a sign wrong in one of the successive transformations (which leads to some sign differences between the two of us).
I'd been carrying out a scaling, which is somehow interacting with the range to cause trouble. I multiplied by the diagonal matrices given by diagonal(s1, 1/s1, 1) and diagonal(1, s2, 1/s2)
These are matrices of norm 1, so they don't change the volume of the unit ball that they're operating on. I believe that I've got a problem with these, however. What I did was replace that by the simpler
diagonal(s1, 1, 1) and diagonal(1, s2, 1), and counted on the range to take care of the third axis. It seems to be working better....

Website maintained by Andy Long. Comments appreciated.