To comment on the following update, log in, then open the issue: http://www.openoffice.org/issues/show_bug.cgi?id=47296
------- Additional comments from [EMAIL PROTECTED] Mon Sep 5 14:27:07 -0700 2005 ------- The reason why the old algorithm failed is because of an overflow in double-precision variable during a calculation of binomial coefficient (note that a calculation of hypergeometric distribution involves calculations of three binomial coefficients). binomial coefficient: (n, k) = n! / k!(n-k)! In the old method, these three factorials n!, k! and (n-k)! were immediately calculated, which caused an overflow. In my proposed method, I expand these factorials into individual elements e.g. 10! -> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 and store them into an enumerator array or a denominator array (whichever is appropriate). These elements get accumulated in these two arrays during all three binomial coefficient calculations. Once that's done, the elements in these two arrays are sorted in descending order (from largest to smallest), and divisions are performed on all stored elements sequentially from the largest elements downward (see my proposed patch for implementation details). This probably is the safest way of calculation when the calculation involves multiple factorial elements. And I recommend the way I calculated the hypergeometric distribution here be applied to other calculation methods that may have a similar numerical overflow problem. Kohei --------------------------------------------------------------------- Please do not reply to this automatically generated notification from Issue Tracker. Please log onto the website and enter your comments. http://qa.openoffice.org/issue_handling/project_issues.html#notification --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
