[R-sig-Geo] Grid neighbours - cell2nb and size of moving window

[Anne Ghisla]

Hello list, and sorry for cross-posting,

I'm calculating Moran's I on a grid dataset, with z value corresponding
to animal relocations' density in each cell.

On an ecological point of view, according to Prof. Baddeley [0], animal
relocations should be considered as trajectories instead of point
patterns. Anyway, most home range estimators treat datasets as point
patterns, and moreover not all datasets have enough autocorrelation to
be handled as trajectories.

This being said, I know that Moran's I is dependent on size and
resolution of the grid, on choice of neighbours and weights. For the
first point, as I'm comparing home range estimators that draw a grid as
intermediate step, I use that grid as it is. For second and third
points, I'd be grateful for advice - at the moment, I use cell2nb with
type=queen and style weight="W".

Is there a more reliable choice of neighbors and/or weights, without
taking into account ecological motivations? I've thought of encompassing
all cells with a command like dnearneigh(coords, d1=0, d2=max(dists))
and weight them with inverse distance, as shown in "Applied Spatial Data
Analysis in R", pp. 247;253.

Thanks in advance for any hint,

Anne Ghisla

[0] http://www.csiro.au/resources/Spatial-Point-Patterns-in-R.html


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: Questa ? una parte del messaggio	firmata digitalmente
URL: <https://stat.ethz.ch/pipermail/r-sig-geo/attachments/20090917/af3a9020/attachment.bin>