r - Correlation/p values of all combinations of all rows of two matrices

Question

I would like to calculate the correlation and the p value of that correlatio of each species (bac) to each of the factors (fac) in a second data frame. Both were measured at the same number of stations, but the number of bac and fac don't match.

bac1 <- c(1,2,3,4,5)
bac2 <- c(2,3,4,5,1)
bac3 <- c(4,5,1,2,3)
bac4 <- c(5,1,2,3,4)
bac <- as.data.frame(cbind(bac1, bac2, bac3, bac4 ))
colnames(bac) <- c("station1", "station2", "station3", "station4")
rownames(bac) <- c("bac1", "bac2", "bac3", "bac4", "bac5")

fac1 <- c(1,2,3,4,5,6)
fac2 <- c(2,3,4,5,1,6)
fac3<- c(3,4,5,1,2,6)
fac4<- c(4,5,1,2,3, 6)
fac <- as.data.frame(cbind(fac1, fac2, fac3, fac4))
colnames(fac) <- c("station1", "station2", "station3", "station4")
rownames(fac) <- c("fac1", "fac2", "fac3", "fac4", "fac5", "fac6")

I imagine the result looking somewhat like this, somewhere keeping the names to know which combination is presented:

bac1-fac1 cor1 p1
bac1-fac2 cor2 p2
bac1-fac3 cor3 p3

bac2-fac1 corx px...

I have looked at function rcorr from Hmist and corr.test from psych, but can't find an example with the neccessary permutation of rows...Any ideas?

Answer 1

If you restructure your data, such that you compute correlation between paired columns, it would be super easy.

tbac <- data.frame(t(bac))
tfac <- data.frame(t(fac))

f <- function (x, y) cor(x, y)

tab <- outer(tfac, tbac, Vectorize(f))

as.data.frame.table(tab)

I had an answer using the same idea: Match data and count number of same value.