matrixStats: Fast column- and row-wise versions of variance coded in C++

matrixStatsR Documentation

Fast column- and row-wise versions of variance coded in C++

Description

These two functions are fast C++ versions for column- and row-wise variance calculation on matrices/data.frames and are meant to substitute the classical apply(mat, 1, var) approach.

Usage

  colVarsC(x)
  rowVarsC(x)  

Arguments

x

a matrix or data.frame

Details

They are coded in a way that they automatically remove NA values, so they behave like na.rm = TRUE; a column/row with fewer than two non-missing values gives NaN. Both functions require x to have at least two columns: for a single column, colVarsC stops with "'x' should have more than one column!" and rowVarsC with "'x' should have more than one column!".

Value

A (named) vector of column/row variances, with the column/row names of x.

Author(s)

Andrej-Nikolai Spiess

Examples

## Speed comparison on a large matrix:
## rowVarsC is much faster than apply().
MAT <- matrix(rnorm(10 * 100000), ncol = 10)
system.time(RES1 <- apply(MAT, 1, var))
system.time(RES2 <- rowVarsC(MAT))
all.equal(RES1, RES2)

propagate documentation built on Nov. 10, 2026, 9:07 a.m.