function r = nacf(y,M)
% This function computes the normalized autocovariance function of a signal y using
% the lagged product algorithm for the biased estimator; maximum lag is
% M time units, r is an M+1 dimensional vector. Remember a large bias is
% induced when M > 0.1N!
r = zeros(M+1,1);
N = length(y);
m = mean(y); var = std(y).^2 *(N-1)/N;
for k = 0:M
y1 = y(1:N-k) - m;
y2 = y(1+k:N) - m;
vecsq = y1 .* y2;
r(k+1) = (sum(vecsq)/N) / var; % covariance/variance
end