TITLE: What's an efficient way to calculate covariance for a large data set?
QUESTION [8 upvotes]: What is the best algorithm for computing covariance that would be accurate for a large number of values like 100,000 or more?

REPLY [2 votes]: A single pass stable algorithm has been discovered in the time since this question has been originally answered:
Bennett, Janine, et al. "Numerically stable, single-pass, parallel statistics algorithms." Cluster Computing and Workshops, 2009. CLUSTER'09. IEEE International Conference on. IEEE, 2009.
An implementation is given in Boost.