[R-sig-Geo] Complete neighbours

German Sanchez Hernandez german.sanchez at ais-int.com
Sun Sep 5 22:11:48 CEST 2010


Dear Roger (and list):

Thank you for your response. The idea to complete the neighbourhood list is to try to "define" some regions using the features of their nearest regions.

I made a modification of union.nb function, once I've learned how to edit the nb objects. I'll take a look on the nblag_cumul function.

Thanks and regards,
German Sanchez

El 05/09/2010, a las 21:14, Roger Bivand escribió:

> On Fri, 3 Sep 2010, German Sanchez Hernandez wrote:
> 
>> Dear list,
>> 
>> I would like to complete my neighbour list in order to increase the number of neighbours of regions with less neighbours (links) than a certain number. The idea is to complete these links using knearneigh function.
>> 
>> 
>> My try:
>> 
>> nb <- poly2nb(map)
>> # Compute distribution of linkage:
>> distr <- as.data.frame(table(card(nb)))
>> # Try to identify the "isolated" regions (equal or less than, for example, 3 neighbours
>> minim <- 3 # It's a computed number
>> isolated <- subset(nb, subset=card(nb) <= minim)
>> # Get the "isolated" region IDs
>> regions <- attr(unclass(isolated), 'region.id')
>> # Modify nb using knearneigh ¿?
> 
> For the edited version of nblag_cumul on R-forge on the spdep project, you'd do something like:
> 
> example(columbus)
> table(card(col.gal.nb))
> k4 <- knn2nb(knearneigh(coordinates(columbus), 4))
> mod_nb <- nblag_cumul(list(col.gal.nb, k4))
> table(card(mod_nb))
> 
> but I'm not at all sure that this is a good idea. Most often observations with few neighbours are on the edge of the data set, so there is a very good reason for their apparent "isolation".
> 
> Hope this helps,
> 
> Roger
> 
>> 
>> 
>> # Alternative 2:
>> nbmat <- nb2mat(nb)
>> # Modify nbmat ¿?, mat2listw, listw2nb
>> 
>> 
>> # Alternative 3:
>> listw <- nb2listw(nb)
>> # Modify listw ¿?, listw2nb
>> 
>> 
>> The question is:
>> 
>> How to modify nb (or nbmat or listw) to increase the number of links of the "isolated" regions?
>> 
>> Any idea?
>> 
>> Thank you!
>> Germán Sánchez
>> _______________________________________________
>> R-sig-Geo mailing list
>> R-sig-Geo at 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.Bivand at nhh.no



More information about the R-sig-Geo mailing list