Talk:Cross-covariance

is this article incorrect: I thought the x-cov of lag L was defined as the /expectation/ of x_t(y_{t-L}.  In particular this means that at the edges of the data we need to adjust for the number of legal datapoints, so take the sum of them and divide by the number of legal points.  (In matlab this is done by xcov 'unbiased' to give the unbiased estimator).  —Preceding unsigned comment added by 143.167.74.60 (talk) 10:09, 26 October 2007 (UTC)


 * The actual mathematical definition is for infinite length sequences, even if the number of non-zero values is finite. Then, mathematically, there is no "edge".  What you (or Matlab) does with a finite-length function is a free choice, not governed by the true definition of correlation.  It's a cop-out for the article to leave the limits of integration specified only as "for the appropriate values of t".


 * An actual error, however, is that article is describing cross-correlation. Cross-covariance is the cross-correlation of each function minus its mean value.
 * --Bob K (talk) 12:00, 23 October 2008 (UTC)

Assessment comment
Substituted at 00:59, 12 June 2016 (UTC)