Raster correlation: Error (arg6) in raster.modified.ttest (R)

Question

I am performing pairwise spatial correlation of 6 rasters using the raster.modified.ttest() function in R, as per this very useful post.

I have consistently had the following error with my data;

Error in SpatialPack::modified.ttest(x.var, y.var, sp::coordinates(x[nb[[i]],  : 
  NA/NaN/Inf in foreign function call (arg 6)

and yet no errors with any dummy data. The following will yield a result;

s.pix1 <- as(raster(xmn=0,xmx=10,ymn=0,ymx=10,res=1,vals=sample(1:5,100,replace=T)), "SpatialGridDataFrame")
s.pix1[sample(1:100,25)]<-NA
s.pix2 <- as(raster(xmn=0,xmx=10,ymn=0,ymx=10,res=1,vals=sample(1:5,100,replace=T)), "SpatialGridDataFrame")
s.pix2[sample(1:100,25)]<-NA

corr <- raster.modifed.ttest(x=s.pix1, y=s.pix2)

plot(raster(corr), main="spatially adjusted raster correlation")

Very small sample areas of my data would also yield results. However forcing larger contiguous areas of NA data into the dummy data recreated the error;

s.pix1 <- as(raster(xmn=0,xmx=10,ymn=0,ymx=10,res=1,vals=sample(1:5,100,replace=T)), "SpatialGridDataFrame")
s.pix1[c(1:5,11:15,21:25,31:35,41:45)]<-NA
s.pix2 <- as(raster(xmn=0,xmx=10,ymn=0,ymx=10,res=1,vals=sample(1:5,100,replace=T)), "SpatialGridDataFrame")
s.pix2[c(1:5,11:15,21:25,31:35,41:45)]<-NA

corr <- raster.modifed.ttest(x=s.pix1, y=s.pix2)

Error in SpatialPack::modified.ttest(x.var, y.var, sp::coordinates(x[nb[[i]],  : 
  NA/NaN/Inf in foreign function call (arg 6)

Could be a bug? @JeffreyEvans provided the excellent answer in the linked post

Answer 1

This is an old post, but I am just sharing a similar experience for whoever will face this type of situation. My error message was:

Error in h(simpleError(msg, call)) : error in evaluating the argument 'x' in selecting a method for function 'raster': NA/NaN/Inf in foreign function call (arg 6)
In addition: Warning message:
In cor(cbind(x, y)) : the standard deviation is zero

Such an error occurred when either one of the rasters had repeating values (NA or identical values) over certain times within a given window. The workaround was to set "d" parameter larger than what "AUTO" option would set.

For instance, my raster had a cell-size of 1000 (meters), and if I used the "Auto" setting d would have been 4242.641. However, I found out that some cases required setting d over 9000 just to avoid crashing.