Skip to content

Is there a way to get cor to work? #90

@ekernf01

Description

@ekernf01

When I try cor, I get errors like Error in cor(DelayedArray(x), DelayedArray(y)) : 'x' must be numeric. Am I missing something simple?

If it's not implemented, I have a little function for this and I could send a pull request.

#' Drop-in replacement for cov or cor; tolerates DelayedArray input. Pearson correlation only. 
#'
delayedCov = function(x, y = x, do_cor = F){
  mux = colMeans(x)
  muy = colMeans(y)
  zx = sweep(x, 2, mux, "-")
  zy = sweep(y, 2, muy, "-")
  covxy = t(zx) %*% zy
  covxy = covxy / nrow(x)
  if(do_cor){
    sx = sqrt(colMeans(zx^2))
    sy = sqrt(colMeans(zy^2))
    covxy = Matrix::Diagonal(x = 1/sx) %*% covxy %*% Matrix::Diagonal(x = 1/sy)
  }
  covxy
}
# tests
x = matrix(rnorm(1000), ncol = 10)
y = matrix(rnorm(1000), ncol = 10)
plot(delayedCov(x, y),  cov(x, y)); abline(a = 0, b = 1)
plot(delayedCov(x, y, do_cor = T),  cor(x, y)); abline(a = 0, b = 1)
plot(delayedCov(DelayedArray(x), DelayedArray(y), do_cor = T),  cor(x, y)); abline(a = 0, b = 1)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions