<meta http-equiv="Content-Type" content="text/html; charset=utf-8"><div dir="auto"><div dir="auto">Hello Roger, thanks for the tips.</div><div dir="auto"><br></div><div dir="auto">In fact, I believe that I will extract the coordinates by the gCentroid function, from the rgeos package.</div><div dir="auto"><br></div><div dir="auto">The results look better.</div><div dir="auto"><br></div><div dir="auto">Also, thank you very much for the GWR tip, in my case, it is just a complementary result, but in fact, is good to know about it.</div><div dir="auto"><br></div><div dir="auto">A hug.</div><div dir="auto"><br></div><div dir="auto">Your tips are helping me so much!</div></div><div class="gmail_extra"><br><div class="gmail_quote">Em 4 de nov de 2020 17:32, Roger Bivand <Roger.Bivand@nhh.no> escreveu:<br type="attribution"><blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><p dir="ltr">On Wed, 4 Nov 2020, Pietro Andre Telatin Paschoalino wrote:
<br>
<br>
> Hello Roger,
<br>
>
<br>
> Yes, I'm using spdep package to construct the matrices.
<br>
>
<br>
<br>
Do not use matrices, use neighbour or weights objects. Dense matrices with
<br>
mostly zero values are a waste of space.
<br>
<br>
> I used the knearneigh function (in which I entered the coordinates), as
<br>
> well as the poly2nb function.
<br>
<br>
Yes, however, you change the support of the aggregated data units by doing
<br>
this, so knn is less well supported than contiguity in general.
<br>
<br>
>
<br>
> I think that I only have one of these rings but the size of the polygons
<br>
> is relatively large (they are political divisions) of a country. In that
<br>
> case, can I keep extracting coordinates by the coordinates function?
<br>
>
<br>
<br>
The centroids are more dependent (and so the distances between point
<br>
representations) on the boundary geometries than are simple contiguities.
<br>
Change of support is always something of a problem. Practically, yes, you
<br>
can do it, but whether it makes sense or not in terms of the data is a
<br>
different question.
<br>
<br>
> I didn't understand why you suggested avoiding GWR. Are you commenting
<br>
> on the methodology itself?
<br>
<br>
Yes, it may be used for exploring to detect possible omitted covariates
<br>
but should not be used for modelling unless that is definitely the data
<br>
generating process. Random data generates patterned (insignificant)
<br>
coefficient maps.
<br>
<br>
Roger
<br>
<br>
>
<br>
> Pietro Andre Telatin Paschoalino
<br>
> Doutorando em Ciências Econômicas da Universidade Estadual de Maringá - PCE.
<br>
>
<br>
>
<br>
>
<br>
<br>
--
<br>
Roger Bivand
<br>
Department of Economics, Norwegian School of Economics,
<br>
Helleveien 30, N-5045 Bergen, Norway.
<br>
voice: +47 55 95 93 55; e-mail: Roger.Bivand@nhh.no
<br>
https://orcid.org/0000-0003-2392-6140
<br>
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en</p>
</blockquote></div><br></div>