Optimizing apply function output

I have the following function:

library (reshape)
phenotype <- rnorm (100)
data <- matrix(rnorm(1000), nrow = 10, ncol=100)

spearman.p <-
             reshape(
                     melt(
                          apply(data, 1, function(y){
                                        cor.test(y,phenotype,method="spearman")
                                                    }[c("p.value", "estimate")]
                               )
                          ), timevar="L2", idvar="L1", direction="wide开发者_如何学JAVA"
                       )

that I would like to know if there is a more efficent way of getting out the p.value and estimate from a "apply"ed cor.test

Can anyone provide some suggestions?

This is the best I can come up with at the moment.

FUN <- function(y) {
  test <- cor.test(y,phenotype,method="spearman")
  out <- unlist(test[c("p.value", "estimate")])
}
t(apply(data, 1, FUN))

This would be more compact and delivers the p.values from the duplicated data. Is that what you wanted?:

 dtt <- do.call(rbind, apply(data, 1, function(y){
                            cor.test(y,phenotype,method="spearman")
                                  }[c("p.value", "estimate")]
                                 ) )
  dtt
 ###      p.value   estimate 
     [1,] 0.2305644 0.1208641
     [2,] 0.2305644 0.1208641
     [3,] 0.2305644 0.1208641
     [4,] 0.2305644 0.1208641
     [5,] 0.2305644 0.1208641
     [6,] 0.2305644 0.1208641
     [7,] 0.2305644 0.1208641
     [8,] 0.2305644 0.1208641
     [9,] 0.2305644 0.1208641
    [10,] 0.2305644 0.1208641

Edit: If you are looking for speed and/or the possibility of easily transporting to parallel-oriented platforms then add this to the list of candidates:

 pmtx <- matrix(NA, nrow=nrow(data), ncol=2)
 for( i in 1:nrow(data) ) {
  pmtx[i, 1:2 ] <- unlist(cor.test(data[i, ], 
                                   phenotype, 
                                   method="spearman")[c("p.value", "estimate")] ) }
 pmtx