R/ff_functions.R

In markheron/nucular: utility functions used for the analysis of genomic nucleosome data

##' function variations to run on ff vectors
##' 
##' Copies of several functions that are originally for numeric vectors, but in this *_ff variations work on ff vectors
##' 
##' @name ff_functions
##' @author Mark Heron
##' 
##' @import ffbase
NULL



##' cov_ff
##' 
##' Computes the covariance between two ff vectors.
##' 
##' @export
##' @param x (numeric ff vector)
##' @param y (numeric ff vector)
##' @param centered (boolean) are the means of x and y == 0?
##' @return covariance between x and y
##' 
cov_ff <- function(x,y,centered=FALSE) {
  
  if(centered) {
    return( sum( x*y , na.rm=TRUE) / (sum( !is.na(x) & !is.na(y) )-1) )
  } else {
    return( cov_ff(x- mean(x, na.rm=TRUE),(y-mean(y, na.rm=TRUE)), centered=TRUE ) )
  }
}



##' sd_ff
##' 
##' Computes the standard deviation of a vector.
##' 
##' @export
##' @param x (numeric ff vector)
##' @param centered (boolean) is the mean of x == 0?
##' @return standard deviation of x
##' 
sd_ff <- function(x, centered=FALSE) {
  
  if(centered) {
    return( sqrt(sum( x^2 , na.rm=TRUE) / (sum(!is.na(x))-1) ) )
  } else {
    return( sd_ff(x-mean(x, na.rm=TRUE), centered=TRUE) )
  }  
}



##' cor_ff
##' 
##' Computes the correlation between two ff vectors.
##' 
##' @export
##' @param x a ff vector
##' @param y a second ff vector
##' @param method one of "pearson" (default) or "spearman", which correlation should be computed. "spearman" has to briefly transform the ff's to vectors for rank.
##' @return correlation between x and y
##' 
cor_ff <- function(x,y, method="pearson") {
  
  if(method == "pearson") {
    centered_x = x - mean(x, na.rm=TRUE)
    centered_y = y - mean(y, na.rm=TRUE)
  } else if(method == "spearman") {
    ranked_x = ff::as.ff(rank(ff::as.ram(x), na.last="keep")) # as.ram has to be used or rank fails!
    ranked_y = ff::as.ff(rank(ff::as.ram(y), na.last="keep"))
    return(cor_ff(ranked_x, ranked_y, method="pearson"))
  } else {
    stop("only 'pearson' and 'spearman' correlations implemented in cor_ff")
  }
  return( cov_ff(centered_x, centered_y, centered=TRUE) / (sd_ff(centered_x, centered=TRUE)*sd_ff(centered_y, centered=TRUE)) )  
}