User talk:Brianboonstra

This C code implements some good and bad ways of calculating correlation between two series, demonstrating the errors arising from less stable implementations.

=Watch it happen= Might I suggest an extension? Suppose X is a set of test values, such as X = {3,1,4,1,5,9,2} or similar - actually, {1,2,3,4,5,6} has merit, as sums, sums of squares, etc. can be calculated mathematically, with ease. For test purposes, rather than trying just (9000000 + X) as above, run the algorithm with (X), then (10 + X), then (100 + X), then (1000 + X), etc. and watch the various calculations fall apart. Try also in single and double and quadruple precision. NickyMcLean (talk) 04:04, 6 April 2009 (UTC)

algorithm -- how did you come to it?
Hi, I believe that you're the author of the algorithm for one pass linear correaltion coefficient? At least you contributed it here? I beleive you referenced "Elements of Statistical Computing: Numerical computation," by Ronald Aaron Thisted, the pages about sweep operator, which I've found (partially) in Google books, still what's there is a lot of theory with matrices I don't even see how that maps to the problem in question (which has two vectors as input). So as far as I understand your algorithm is still something untrivial: for example "Numerical Recepies" book does 2 passes. They keep stability by using differencies but they didn't reach elegancy you showed. Is there a chance that you write somewhere else (a you'd have problems adding it to the wikipedia article directly) in more details about that particular algorithm? Thank you. Janjs (talk) 17:43, 17 June 2009 (UTC)


 * I'll try to write something up. For now, I should just clarify the translation of all that matrix algebra to this algorithm.  Computing correlation is a degenerate case of the use of the sweep operator.  So if you implement something like the formulas found in Thisted (many books have equivalents) and then eliminate all the unused bits, this is what you end up with.Brianboonstra (talk) 13:47, 18 August 2009 (UTC)

Unblock request
The IP is from Forcepoint Cloud Services and is hardblocked. — Berean Hunter   (talk)  15:29, 18 April 2018 (UTC)

Women in Red World Contest
Hi. We're into the last five days of the Women in Red World Contest. There's a new bonus prize of $200 worth of books of your choice to win for creating the most new women biographies between 0:00 on the 26th and 23:59 on 30th November. If you've been contributing to the contest, thank you for your support, we've produced over 2000 articles. If you haven't contributed yet, we would appreciate you taking the time to add entries to our articles achievements list by the end of the month. Thank you, and if participating, good luck with the finale!