系数和相关的成对组合p.value？

Question

I want to compute the pairwise combinations of coefficient and correlation pvalue for a matrix by columns.

Here I use two functions:

allCoef<- function(Y,X) {  lm(Y~X+0)$coef }
allCorr.p<- function(Y,X) {  cor.test(Y,X)$p.value }

For example I have a matrix of A:

A= matrix(sample(1:100,16),4,4)
apply(Y=A,2,allCoef,X=A)

works fine.

apply(Y=A,2,allCorr.p,X=A)

However, shows Error in cor.test.default(Y, X) : 'x' and 'y' must have the same length . Can somebody please advise what have I done wrong here? I am using the same matrix so the length of columns should be identical.

Answer 1

You can use the combn function to generate all combinations of column comparisons and then apply across this matrix using cor.test on the combinations of columns of A (this assumes A is available in your global environment):

# All combinations of pairwise comparisons
cols <- t( combn(1:4,2) )
apply( cols , 1 , function(x) cor.test( A[,x[1] ] , A[ , x[2] ] )$p.value )
#[1] 0.9893876 0.9844555 0.5461623 0.7987615 0.7414658 0.1061751

The pairwise combinations of columns generated by the combn function is:

     [,1] [,2]
[1,]    1    2
[2,]    1    3
[3,]    1    4
[4,]    2    3
[5,]    2    4
[6,]    3    4

Your apply(Y=A,2,allCorr.p,X=A) did not work as expected because (disregarding that you do not need to use Y=A ) you pass the whole matrix as the second argument to your function, so X actually has the length of all columns in your matrix.