Efficiently perform row-wise distribution test

Question

I have a matrix in which each row is a sample from a distribution. I want to do a rolling comparison of the distributions using ks.test and save the test statis

Answer 1

A quick and dirty implementation in Rcpp

// [[Rcpp::depends(RcppArmadillo)]]
#include  

double KS(arma::colvec x, arma::colvec y) {
  int n = x.n_rows;
  arma::colvec w = join_cols(x, y);
  arma::uvec z = arma::sort_index(w);
  w.fill(-1); w.elem( find(z <= n-1) ).ones();
  return max(abs(cumsum(w)))/n;
}
// [[Rcpp::export]]
Rcpp::NumericVector K_S(arma::mat mt) {
  int n = mt.n_cols; 
  Rcpp::NumericVector results(n);
  for (int i=1; i

for matrix of size (400, 30000), it completes under 1s.

system.time(K_S(t(mt)))[3]
#elapsed 
#   0.98 

And the result seems to be accurate.

set.seed(1942)
mt <- matrix(rnorm(400*30000), nrow=30000)
results <- rep(0, nrow(mt))
for (i in 2 : nrow(mt)) {
  results[i] <- ks.test(x = mt[i - 1, ], y = mt[i, ])$statistic
}
result <- K_S(t(mt))
all.equal(result, results)
#[1] TRUE