[R-sig-Geo] Producing a conditional adjacency matrix

[Roger Bivand]

On Tue, 1 Oct 2019, Rhys Dubin wrote:

> Hi Roger,
>
> Including a bit more detail below (my apologies if this is just complete gibberish — first time posting to a forum like this):
>
> First, I add a random ID variable to the columbus dataset indicating the variable I’d like to base my exclusions on in the final neighbor list:
>> columbus$ID <- sample(1:3, nrow(columbus), replace=T)
>
> In my case, that produced the following IDs for the first 4 polyids:
>
> POLYID 1: 3
> POLYID 2: 2
> POLYID 3: 3
> POLYID 4: 1
>
> Second, col.ga.nb indicates that each of the previous POLYID’s has the following neighbors
>
>> View(col.gal.nb)
>
> POLYID 1 (w new assigned ID of 3): 2, 3
> POLYID 2 (“ “ 2): 1, 3, 4
> POLYID 3 (“ “ 3): 1, 2, 4, 5
> POLYID 4 (“ “ 1): 2, 3, 5, 8
>
> Third (or brevity’s sake) looking only at POLYID 1 (with new ID = 3):
>
> Its neighbors from col.ga.nb (POLYID 2 and 3) have ID values of 3 and 1 respectively
>
> My end goal would be to produce a new list in which POLYID 1’s only neighbor is POLYID 3 (with an ID of 1). POLYID 2 would be excluded, as its ID value matches its parent.

Leave col.gal.nb as it is, and generate a list of general weights:

> library(spdep)
Loading required package: sp
Loading required package: spData
Loading required package: sf
Linking to GEOS 3.7.2, GDAL 3.0.1, PROJ 6.2.0
> data(columbus, package="spData")
> ls()
[1] "bbs"        "col.gal.nb" "columbus"   "coords"     "polys"
> col.gal.nb[[1]]
[1] 2 3
> set.seed(1)
> columbus$ID <- sample(1:3, nrow(columbus), replace=T)
> columbus$ID[1]
[1] 1
> columbus$ID[col.gal.nb[[1]]]
[1] 3 1 # second entry equal
> glist <- lapply(seq(along=col.gal.nb), function(i)
   as.integer(columbus$ID[i] != columbus$ID[col.gal.nb[[i]]]))
> glist[[1]]
[1] 1 0 # second weight zero

From there use nb2listw() to generate weights if that is what you need. 
lapply() is always worth learning - do not use View() on objects that are 
not data.frames or matrices, it either tells nothing or is misleading.

Hope this clarifies,

Roger

>
> The same would be true of each subsequent polygon.
>
> In essence, the same format you mentioned at the end of your reply: a 
> glist including all neighbors j with a value of x not equal to x[i], and 
> excluding all with x[j] == x[i].
>
> Best,
>
> Rhys
>
>
>> On Oct 1, 2019, at 1:22 PM, Roger Bivand <Roger.Bivand using nhh.no> wrote:
>>
>> On Mon, 30 Sep 2019, Rhys Dubin wrote:
>>
>>> Hi all,
>>>
>>> I have a question regarding matrices (produced using poy2nb in spdep).
>>>
>>> I’m trying to produce an adjacency matrix from a GIS Polygon shapefile with about 500 individual polygons (queen or rook — it doesn’t really matter for now), but conditional on the value of the parent polygon.
>>>
>>> For instance, if a polygon had an arbitrary value of 1, I want to return a list of all neighboring polygons that have a value of 2 or 3 (ie. ≠ 1). The same would go for a polygon with a value of 2 (ie. return all neighbors with a value of ≠ 2).
>>>
>>> Is there a solution to this issue built into spdep, do I need to use a different package, or is it possible to accomplish this with a subsetting command?
>>
>> No built-in solution, but to make progress a clear, reproducible example is essential, otherwise we're guessing what you want. For example, add a doctored variable to the eire or columbus data sets to show your needs. Do you want a glist including (1) all neighbours j with a value of x not equal to x[i], and excluding (0) all with x[j] == x[i]?
>>
>> Roger
>>
>>>
>>> Thanks!
>>>
>>> Rhys
>>>
>>>
>>> _______________________________________________
>>> R-sig-Geo mailing list
>>> R-sig-Geo using r-project.org
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>
>> --
>> Roger Bivand
>> Department of Economics, Norwegian School of Economics,
>> Helleveien 30, N-5045 Bergen, Norway.
>> voice: +47 55 95 93 55; e-mail: Roger.Bivand using nhh.no
>> https://orcid.org/0000-0003-2392-6140
>> https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en
>
>

-- 
Roger Bivand
Department of Economics, Norwegian School of Economics,
Helleveien 30, N-5045 Bergen, Norway.
voice: +47 55 95 93 55; e-mail: Roger.Bivand using nhh.no
https://orcid.org/0000-0003-2392-6140
https://scholar.google.no/citations?user=AWeghB0AAAAJ&hl=en