Dear Rusers,
I’m tryingto figure out what I think is a pretty simple thing for anyone who
knows about correlograms.I’ve a regular grid (say 5*5 points) with some
quantity associated to eachpoint (count data). I’m trying to verify whether
this quantity is regularly /randomly or “clusterdly” distributed on the grid.
I’ve decided to give a shotto the sp.correlogram {spdep}.
I first createda grid using cell2nb:
grid <- cell2nb(5,5)xyc <- attr(grid,"region.id")xy
<-matrix(as.integer(unlist(strsplit(xyc, ":"))), ncol=2,
byrow=TRUE)plot(grid,xy) >gridNeighbour list object:Number of regions: 25
Number of nonzero links: 80 Percentage nonzero weights: 12.8 Average number of
links: 3.2 I then usedsp.correlogram, and specified “order = 4” since I figured
the maximum lagbetween 2 points on a 5 by 5 grid is 4… In sp.correlogram we do
not have tospecify a “style” as in moran.test, not sure why so far… anyway.
results<- sp.correlogram(grid, data$quantity, order=4, method = "I")
print(results,"bonferroni") In the “print” tbale, the count ofobservation per
lag order (in brakets) is 25 for each lag. This is what I donot understand,
should not this count be changing with lags? I mean, when you look at the
graph of “grid” Iwould have expected a lower number for lag 4 (say only 15
pairs of observationare “that far”) and a way higher number for lag 1… Does
that make sens toanyone? regards
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
