On Wed, 4 Nov 2020, Pietro Andre Telatin Paschoalino wrote:
> Hello Roger,
>
> Yes, I'm using spdep package to construct the matrices.
>
Do not use matrices, use neighbour or weights objects. Dense matrices with
mostly zero values are a waste of space.
> I used the knearneigh function (in which I entered the coordinates), as
> well as the poly2nb function.
Yes, however, you change the support of the aggregated data units by doing
this, so knn is less well supported than contiguity in general.
>
> I think that I only have one of these rings but the size of the polygons
> is relatively large (they are political divisions) of a country. In that
> case, can I keep extracting coordinates by the coordinates function?
>
The centroids are more dependent (and so the distances between point
representations) on the boundary geometries than are simple contiguities.
Change of support is always something of a problem. Practically, yes, you
can do it, but whether it makes sense or not in terms of the data is a
different question.
> I didn't understand why you suggested avoiding GWR. Are you commenting
> on the methodology itself?
Yes, it may be used for exploring to detect possible omitted covariates
but should not be used for modelling unless that is definitely the data
generating process. Random data generates patterned (insignificant)
coefficient maps.
Roger
>
> Pietro Andre Telatin Paschoalino
> Doutorando em Ciências Econômicas da Universidade Estadual de Maringá - PCE.
>
>
>
--
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; e-mail: roger.biv...@nhh.no
https://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en
_______________________________________________ R-sig-Geo mailing list R-sig-Geo@r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
