Question

I have to make a chi square test of two physical variables with R. I’am trying with:

Library('MASS')
Library('gplots')
data<-read.table('data.dat',head=F) 
pp<-hist2d(data$V2,data$V3)
chisq.test(pp$counts)

but R says me:

Pearson's Chi-squared test

    data:  pp$counts 
    X-squared = NaN, df = 240, p-value = NA

I have used this script in past to performs chi square but now it does not work. where is the problem?

Answer 1

hist2d bins the data, but if some of the bins are always empty,
the chi squared statistic is not defined (because of a division by zero).
You can try to reduce the number of bins,
or discard the bins that are empty.

library(gplots)
d <- data.frame( rnorm(100), rnorm(100) )

# Discard empty bins
p <- hist2d(d)
i <- apply( p$counts, 1, sum ) > 0
j <- apply( p$counts, 2, sum ) > 0
chisq.test( p$counts[i,j] )

# Reduce the number of bins
p <- hist2d(d,nbins=5)
chisq.test( p$counts )

(From a statistical point of view,
I am not sure that what you are doing is optimal.)