corLocal: Local correlation coefficient

In rspatial/raster: Geographic Data Analysis and Modeling

Local correlation coefficient

Description

Local correlation coefficient for two RasterLayer objects (using a focal neighborhood) or for two RasterStack or Brick objects (with the same number of layers (> 2))

Usage

## S4 method for signature 'RasterLayer,RasterLayer'
corLocal(x, y, ngb=5, 
     method=c("pearson", "kendall", "spearman"), test=FALSE, filename='', ...)


## S4 method for signature 'RasterStackBrick,RasterStackBrick'
corLocal(x, y,  
     method=c("pearson", "kendall", "spearman"), test=FALSE, filename='', ...)

Arguments

x	RasterLayer or RasterStack/RasterBrick
y	object of the same class as x, and with the same number of layers
ngb	neighborhood size. Either a single integer or a vector of two integers c(nrow, ncol)
method	character indicating which correlation coefficient is to be used. One of "pearson", "kendall", or "spearman"
test	logical. If TRUE, return a p-value
filename	character. Output filename (optional)
...	additional arguments as for writeRaster

Value

RasterLayer

Note

NA values are omitted

See Also

cor, cor.test

Examples

b <- stack(system.file("external/rlogo.grd", package="raster"))
b <- aggregate(b, 2, mean)

set.seed(0)
b[[2]] <- flip(b[[2]], 'y') + runif(ncell(b))
b[[1]] <- b[[1]] + runif(ncell(b))

x <- corLocal(b[[1]], b[[2]], test=TRUE )
# plot(x)

# only cells where the p-value < 0.1
xm <- mask(x[[1]], x[[2]] < 0.1, maskvalue=FALSE)
plot(xm)


# for global correlation, use the cor function
x <- as.matrix(b)
cor(x, method="spearman")
 
# use sampleRegular for large datasets
x <- sampleRegular(b, 1000)
cor.test(x[,1], x[,2])

# RasterStack or Brick objects
y <- corLocal(b, flip(b, 'y'))

rspatial/raster documentation built on Oct. 12, 2024, 10:49 a.m.