Thanks for the kind comments, Don. You are correct. I have a similar spreadsheet to show that the AXP BA pair -- (do not trade this at home, BA is not a bank and AXP does not build airplanes so they would make a lousy pair) -- does, in fact have the correct correlation on the www.pairtrader.com website. We are not immune to errors, but in this case, it is just a matter of being aware of the parameters used to construct the correlations. As one of the designers of that website, I apologize that we failed to include a detailed explanation of how the correlation is being calculated. The idea of that page is just to show an example of the kind of statistics that are available to pairtraders on the site. In the site (which is available for a two week free trial), you can get all kinds of additional statistics and charts including a chart of how correlations vary with time for any of hundreds of thousands of pairs (only about 16,000 of which make any sense to trade). For some obscure reason people tend to obsess about this correlation issue, so we will probably just take that number out of the free sample statistics page to avoid confusion.
so basically they just average the HLC of the day and use that data on the correlation.... thanks for the spreadsheet, don. perhaps mysticman or someone else can explain the advantage or reason to use logarithms such as mystic man does ? thanks.
You use logarithms so that the measures make sense over time as the average value drifts around. log(100)-log(50) == log(50)-log(25) 100/50 == 50/25 exp(log(100)-log(50)) ==100/50