spatial clusters

Dear all,

Finally, with your help, I make some progress to apply SKATER algorithm to
European countries.

Here is the R code, following the example from  spdep package pdf:

country<-wrld_simpl[wrld_simpl$ISO3%in%c("AZE", "ALB", "ARM", "BIH",

"BGR", "CYP", "DNK", "IRL", "EST", "AUT", "CZE", "FIN",  "FRA", "GEO",
"DEU", "GRC", "HRV", "HUN", "ISL", "ITA", "LVA", "BLR", "LTU", "SVK",
"LIE", "MLT", "BEL", "AND",  "LUX", "MCO", "NLD", "NOR", "POL", "PRT",
"ROU", "MDA", "RUS", "SVN"),]

ske<-spCbind(country, rclasif)
ske at data
de <- data.frame(scale(ske at data[13]))   #col 13 contains the health

indicator (from rclasif)

nbe <- poly2nb(country)
ce <- nbcosts(nbe, de)
le <- nb2listw(nbe, ce, style="B", zero.policy=TRUE)

Error in nb2listw(nbe, ce, style = "B", zero.policy = TRUE) :
neighbours and glist do not conform

Using your corrected list of countries, I see two/three problems:

1) some proximate polygons do not touch because of line generalisation in 
the simplification of the country polygons, use:

nbe <- poly2nb(country, snap=0.1)

to force neighbours to touch.

2) you are left with three real islands - Iceland, Cyprus, and Malta. To 
join them in, consider:

knn4 <- knn2nb(knearneigh(crds, 4, longlat=TRUE),
   row.names=row.names(country))
knn4s <- make.sym.nb(knn4)
nbe1 <- union.nb(nbe, knn4s)

3) nbcosts() does not like single-column data frames, as it is supposed to 
be for multivariate data, so:

de <- data.frame(rnorm(46))
ce <- nbcosts(nbe1, de)
which(sapply(ce, function(x) isTRUE(any(is.na(x)))))

gives an NA at the end of each line of cost distances, but:

de <- data.frame(rnorm(46), rnorm(46))
ce <- nbcosts(nbe1, de)
which(sapply(ce, function(x) isTRUE(any(is.na(x)))))
le <- nb2listw(nbe1, ce, style="B")

doesn't, nor does:

de <- data.frame(rnorm(46), rep(1, 46))
ce <- nbcosts(nbe1, de)
which(sapply(ce, function(x) isTRUE(any(is.na(x)))))
le <- nb2listw(nbe1, ce, style="B")