Hi Roger,
Thank you very much for your quick reply and for creating the spdep package. I
tried the tweaks to the code as you note below, but I still get a warning
message. I have reposted my code below with a few more questions:
> #making inverse distance weights
> neigh.dist <- nbdists(neigh.nb, coords, longlat=TRUE)
> inverse <- lapply(neigh.dist, function(x) (1/(x^2)))
>
> #creating row standardized spatial weights from list
> neigh.listw <- nb2listw(neigh.nb, glist=inverse, style="W", zero.policy=TRUE)
Warning message:
In nb2listw(neigh.nb, glist = inverse, style = "W", zero.policy = TRUE) :
zero sum general weights
> warning()
Warning message:
There was nothing listed here. Is this warning a problem? I was hesitant
about using glist since I didn't think it could be incorporated in the
errorsarlm models (hence the error in the code below). Would using spautolm()
help to get around my error below in the model?
> #making crime counts into rates
> assaul_r <- (assaul/gpop00)*1000
>
Good catch on this one. I meant to make my crime counts into rates. I haven't
found a package where I can model crime counts as Poisson or negative binomial
regressions. Any suggestions?
> #SAR MODEL
> ass05 <- errorsarlm(assaul_r ~ gage2900 + gblack00 + gcrwd00 + gethhet00 +
> ghhincsdln00 + glatino00 + gocc00+ gowner00 +
> gpov00,data=buffalo,listw=neigh.listw, zero.policy=TRUE)
Error in subset.listw(listw, subset, zero.policy = zero.policy) :
Not yet able to subset general weights lists
> traceback()
4: stop("Not yet able to subset general weights lists")
3: subset.listw(listw, subset, zero.policy = zero.policy)
2: subset(listw, subset, zero.policy = zero.policy)
1: errorsarlm(assaul_r ~ gage2900 + gblack00 + gcrwd00 + gethhet00 +
ghhincsdln00 + glatino00 + gocc00 + gowner00 + gpov00, data = buffalo,
listw = neigh.listw, zero.policy = TRUE)
Thanks again for your help!
Adam
Adam Boessen
Doctoral Student
Department of Criminology, Law and Society
University of California, Irvine
aboes...@uci.edu
On Feb 21, 2010, at 11:38 PM, Roger Bivand wrote:
> On Sun, 21 Feb 2010, Adam Boessen wrote:
>
>> My apologies for the duplicate list, but I forgot to use plain text in my
>> last post, and thus there was a problem posting to archive. Here is the
>> post now:
>
> Thanks for cooperating about this, keeping to plain text really helps!
>
>>
>> Hi Everyone,
>> I'm new to R, so thank you in advance for your patience. I'm using US
>> census block groups in Buffalo New York to examine how neighborhood
>> characteristics affect crime. I would like to use an inverse distance
>> weights (distance decay) of block group centroids that is banded at 1
>> kilometer. In other words, I would like to create a one kilometer buffer
>> around each of the centroids, then use row standardized inverse distance
>> weights. Finally, I would like to run a spatial error model using these
>> weights.
>>
>> Here is my code:
>>> library(foreign)
>>> library(spdep)
>>>
>>> buffalo <- read.dta("buffalo_blkgrps.dta")
>>> attach(buffalo)
>>> names(buffalo)
>> [1] "bgidfp00" "gage2900" "gavghhinc00" "gblack00" "gcrwd00"
>> [6] "gethhet00" "ghhincsdln00" "glatino00" "gocc00" "gowner00"
>> [11] "gpop00" "gpov00" "assaul" "robber" "burglr"
>> [16] "motveh" "murder" "larcen" "x" "y"
>>>
>>> #distance based neighbors - making a neighbor list - bounded to 1 kilometer
>>> coords <-(cbind(x,y))
>>> neigh.nb <- dnearneigh(coords, 0, 1, longlat=TRUE)
>
> I'm assuming that you know definitely that your coordinates are geographical
> not projected - I've seen users adding longlat=TRUE when not appropriate.
>
>>> summary(neigh.nb)
>> Neighbour list object:
>> Number of regions: 409
>> Number of nonzero links: 5298
>> Percentage nonzero weights: 3.167126
>> Average number of links: 12.95355
>> 1 region with no links:
>> 12
>> Link number distribution:
>>
>> 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25
>> 26
>> 1 1 3 3 4 5 14 25 28 23 19 26 32 44 29 24 29 25 24 13 15 12 3 4 1 1
>> 1
>> 1 least connected region:
>> 3 with 1 link
>> 1 most connected region:
>> 382 with 26 links
>>>
>>> #making inverse distance weights
>>> neigh.dist <- nbdists(neigh.nb, coords, longlat=TRUE)
>>> inverse <- lapply(neigh.dist, function(x) (1/(x^2)))
>>>
>>> #creating row standardized spatial weights from list
>>> neigh.listw <- nb2listw(inverse, style="W",zero.policy=TRUE)
>
> No, use:
>
> neigh.listw <- nb2listw(neigh.nb, glist=inverse, style="W",
> zero.policy=TRUE)
>
> You pass the list of weights through the glist= argument, as in the example
> for col.w.d in ?nb2listw.
>
>> Error in nb2listw(inverse, style = "W", zero.policy = TRUE) :
>> Not a neighbours list
>>> traceback()
>> 2: stop("Not a neighbours list")
>> 1: nb2listw(inverse, style = "W", zero.policy = TRUE)
>>>
>>>
>>> assault <- errorsarlm(assaul_r ~ gage2900 + gblack00 + gcrwd00 + gethhet00
>>> + ghhincsdln00 + glatino00 + gocc00+ gowner00 + gpov00,
>> + data=buffalo,listw=neigh.listw, zero.policy=TRUE)
>> Error in eval(expr, envir, enclos) : object 'assaul_r' not found
>>
>
> Well, there is no assaul_r in buffalo, is there?
>
> Hope this helps,
>
> Roger
>
>> The model will run when I do not include the code with the inverse weights
>> and only interpoint distances, but it's unclear to me why I can't include
>> the inverse distances when using nb2listw. Any ideas on why this is
>> occurring and/or help to alleviate this issue would be much appreciated.
>> The spatial error model from errorsarlm (package=spdep) will also run fine
>> when I don't include the inverse weights. Is there a better way to go about
>> running this spatial error model using row standardized inverse distance
>> weights? Would spautolm from the spdep package be better suited for these
>> data? Any comments or suggestions are welcome.
>> Thank you for your time!
>>
>> Adam
>>
>> Adam Boessen
>> Doctoral Student
>> Department of Criminology, Law and Society
>> University of California, Irvine
>> aboes...@uci.edu
>>
>> p.s. If it's helpful, here is my version of R:
>>> version
>> _
>> platform i386-apple-darwin9.8.0
>> arch i386
>> os darwin9.8.0
>> system i386, darwin9.8.0
>> status
>> major 2
>> minor 10.1
>> year 2009
>> month 12
>> day 14
>> svn rev 50720
>> language R
>> version.string R version 2.10.1 (2009-12-14)
>>
>> _______________________________________________
>> R-sig-Geo mailing list
>> R-sig-Geo@stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>
>
> --
> Roger Bivand
> Economic Geography Section, Department of Economics, Norwegian School of
> Economics and Business Administration, Helleveien 30, N-5045 Bergen,
> Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
> e-mail: roger.biv...@nhh.no
>
_______________________________________________
R-sig-Geo mailing list
R-sig-Geo@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
