Drop regions with no links?

[thomashudsonuk]

When running poly2nb on a sf spatial dataframe I get regions with no links. I'm using this neighbours list to create matrix for an INLA spatial model, which as I understand it requires no regions with no links. Is there a way to automatically drop those regions with no links/ extract them list to a different variables?

I'm imagine the solution to look something like:

nb <- poly2nb(polygons)
nb$no_links # to display those regions with no links so I can iterate and remove them

or

nb <- poly2nb(polygons, force_no_links = TRUE) # some parameter which does the above automatically, default to FALSE

Is this possible? Please let me know if you need more information.

[rsbivand]

Please see: https://r-spatial.github.io/spdep/articles/subgraphs.html. The main problem is to ensure that all are linked and that there are no separate subgraphs, while keeping the data and the spatial weights properly ordered. The user needs to consider why the observations are not found to be contiguous, which might well be because the polygons are not defined adequately, so automating dropping apparently not linked observations might omit observations that should have been included. Read also https://r-spatial.github.io/spdep/reference/poly2nb.html about the snap argument. The sfislands https://cran.r-project.org/package=sfislands, https://horankev.github.io/sfislands/, tries to automate this, but I think fails by taking too many decisions out of the hands of the user. For INLA also see https://doi.org/10.1016/j.sste.2018.04.002.