[R-sig-Geo] Adding a Few Neighbour Relationships to a nb List

Thu Aug 8 13:52:25 CEST 2019

On Thu, 8 Aug 2019, Stuart Reece wrote:

> Thanks Roger.
> Happy to do as you have requested.
> Which server do you suggest??

You gave the link to the map as:

https://www.samhsa.gov/data/report/2014-2016-nsduh-substate-region-shapefile 
leading to 
https://www.samhsa.gov/data/sites/default/files/cbhsq-reports/NSDUHsubstateShapeFile2016/ShapeFile2016.zip

I guess that is the input map. Produc is built into the plm package. So 
all we need is a script. Your data are no help; creating a reproducible 
example very often guides the person with the problem to a resolution even 
without external help. I see that your map has 300+ entities - in that 
case find another map that matches the setting (US with Alaska and Hawaii 
states, drop to 48 to match Produc).

Roger

> Thanks again,
> Stuart.
>
> -----Original Message-----
> From: Roger Bivand <Roger.Bivand using nhh.no>
> Sent: Thursday, 8 August, 2019 9:35 PM
> To: Stuart Reece <stuart.reece using bigpond.com>
> Cc: 'Dr Stuart Reece' <asreece using bigpond.net.au>; 'Barry Rowlingson' <b.rowlingson using gmail.com>; 'R-sig-geo Mailing List' <r-sig-geo using r-project.org>
> Subject: RE: [R-sig-Geo] Adding a Few Neighbour Relationships to a nb List
>
> On Thu, 8 Aug 2019, Stuart Reece wrote:
>
>> Thankyou for this advice Roger.
>> Happy to provide my work for your review but I am not sure how.
>> This list server has a limit of 50KB and these files are about 100MB....
>> Love to assist but I would need advice on how best to proceed....???
>
> Choose a built-in data set, such as Produc, change two states to ALASKA and HAWAII, drop the ones you chose from the map and any other states not on your map, and use that data set, putting the script needed at the head of your example. Then create the weights, initially with no links to ALASKA and HAWAII, and run the model. Next, edit the neighbour object in script form, and try again. Then anyone with your map object (you gave a link), your script, and the plm and splm packages can reproduce your problem.
>
> Do also run traceback() after errors, and post what you see.
>
> Roger
>
>> Many thanks again,
>> Stuart.
>>
>> -----Original Message-----
>> From: Roger Bivand <Roger.Bivand using nhh.no>
>> Sent: Thursday, 8 August, 2019 9:10 PM
>> To: Stuart Reece <stuart.reece using bigpond.com>
>> Cc: 'Dr Stuart Reece' <asreece using bigpond.net.au>; 'Barry Rowlingson'
>> <b.rowlingson using gmail.com>; 'R-sig-geo Mailing List'
>> <r-sig-geo using r-project.org>
>> Subject: Re: [R-sig-Geo] Adding a Few Neighbour Relationships to a nb
>> List
>>
>> On Thu, 8 Aug 2019, Stuart Reece wrote:
>>
>>> Thankyou so much Vijay and Barry.
>>
>> Yes, thanks!
>>
>>>
>>> Yes I have found that using these techniques I can add nb
>>> relationships nicely.
>>>
>>> And they card() nicely too.
>>>
>>> And when I nb2listw they transform well and when I test them
>>> can.be.simmed(x) they pass this test also.
>>>
>>> When I draw maps with them Like Roger Bivand showed in his examples
>>> for poly2nb – the maps draw just perfectly.
>>
>> This is flying on hope not knowledge. If the assigned values are not
>> within the 1:n set, there will be trouble - this is an S3 class with
>> no validity check. Further, there may be trouble if the length of
>> attr(.,
>> "region.id") is incorrect, but this cannot be checked unless you run
>> traceback() after the error in spgm(). We still need a fully reproducible example ...
>>
>> Roger
>>
>>> However when I use them in the spml or spgm regressions in package
>>> splm they fail with an error message usually about indexes being out
>>> of bounds.
>>>
>>> Running this code ….
>>>
>>> summary(spgm(PC1MI ~ PC1Rx * log(mrjmon),
>>> +              data=MIdf9dfF2, listw=MIdf9sflww,
>>> +              lag=TRUE, moments="fullweights", method="g2sls",
>>> +              model="random", spatial.error=TRUE))
>>>
>>> Generates this error:
>>>
>>> Error in x[, ii] : subscript out of bounds
>>>
>>> I feel that I must be missing something here but am not able to put
>>> my finger on what it is???
>>>
>>> Thanks so much again,
>>>
>>> Stuart.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> From: Dr Stuart Reece <asreece using bigpond.net.au>
>>> Sent: Thursday, 8 August, 2019 2:01 PM
>>> To: 'Barry Rowlingson' <b.rowlingson using gmail.com>; 'Stuart Reece'
>>> <stuart.reece using bigpond.com>
>>> Cc: 'Vijay Lulla' <vijaylulla using gmail.com>; 'R-sig-geo Mailing List'
>>> <r-sig-geo using r-project.org>
>>> Subject: RE: [R-sig-Geo] Adding a Few Neighbour Relationships to a nb
>>> List
>>> Importance: High
>>>
>>>
>>>
>>> Thanks Barry.
>>>
>>> That is so perfect and so super helpful!!!
>>>
>>> And if I want to add three areas to an area – say I want to add 6,7 and 8 to the area 10??
>>>
>>> Please may I have the syntax for that to avoid the integer error??
>>>
>>> Is this also the root of the error about not being the correct index??
>>>
>>> Many thanks again,
>>>
>>> Stuart.
>>>
>>>
>>>
>>> From: Barry Rowlingson [mailto:b.rowlingson using gmail.com]
>>> Sent: Thursday, 8 August 2019 8:51 AM
>>> To: Stuart Reece
>>> Cc: Vijay Lulla; R-sig-geo Mailing List; Stuart Reece
>>> Subject: Re: [R-sig-Geo] Adding a Few Neighbour Relationships to a nb
>>> List
>>>
>>>
>>>
>>> I recently answered a similar question on Stack Overflow where someone needed to add detached polygons to their connected network by connecting them to their nearest neighbour:
>>>
>>>
>>>
>>> https://stackoverflow.com/questions/57269254/how-to-impute-missing-ne
>>> i
>>> ghbours-of-a-spatial-weight-matrix-queen-contiguity/57378930?noredire
>>> c
>>> t=1#comment101246065_57378930
>>>
>>>
>>>
>>> in short, you can treat a `nb` object like a list of vectors: nb[[i]]
>>> is a vector of indexes of objects connected to object `i`
>>>
>>>
>>>
>>> BUT you have to make sure you store integers:
>>>
>>>
>>>
>>> Here's a `nb` object from that question which in summary has this manyneighbours for each region:
>>>
>>>
>>>
>>>> card(nb)
>>> [1] 2 3 4 3 2 0 0
>>>
>>>
>>>
>>> lets set the 6th feature to be a neighbour of the first:
>>>
>>>
>>>
>>>> nb[[6]] = 1
>>>
>>>
>>>
>>> then uh-oh...
>>>
>>>
>>>
>>>> card(nb)
>>> Error in card(nb) :
>>>  INTEGER() can only be applied to a 'integer', not a 'double'
>>>
>>>
>>>
>>> same again only `as.integer`:
>>>
>>>
>>>
>>>> nb[[6]] = as.integer(1)
>>>
>>>
>>>
>>> and its happy:
>>>
>>>
>>>
>>>> card(nb)
>>> [1] 2 3 4 3 2 1 0
>>>
>>>
>>>
>>> if you want to set the nighbours of 6 to several features:
>>>
>>>
>>>
>>>> nb[[6]] = as.integer(c(1,2,3))
>>>> card(nb)
>>> [1] 2 3 4 3 2 3 0
>>>
>>>
>>>
>>> Barry
>>>
>>>
>>>
>>>
>>>
>>>
>>> 	[[alternative HTML version deleted]]
>>>
>>> _______________________________________________
>>> 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
>
>

-- 
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