DO NOT REPLY TO THIS EMAIL, BUT PLEASE POST YOUR BUG· RELATED COMMENTS THROUGH THE WEB INTERFACE AVAILABLE AT <http://issues.apache.org/bugzilla/show_bug.cgi?id=36215>. ANY REPLY MADE TO THIS MESSAGE WILL NOT BE COLLECTED AND· INSERTED IN THE BUG DATABASE.
http://issues.apache.org/bugzilla/show_bug.cgi?id=36215 ------- Additional Comments From [EMAIL PROTECTED] 2005-08-17 21:18 ------- I think you're right. it is coming from the accumulation of numerical error. but I think the solution should be more complicated than a simple short circuit. let's see an example again: if we calculate cumulativeProbability(50) for (6000, 200, 100), the upper tail (X>=50) probability will be 1 - cumulativeProbability(50) = 3.37E-11. but if you do a sum of the point probabilities using probability(50),(51),...,(100), which by textbook definition defines the upper tail, you'll get 6.02E-49. the discrepency, coming from the accumulation of numerical error, results in a false estimate of the real cumulative probability, by a factor of E38. true, both are very small, but in some cases it would matter for sure. that's why my suggestion is, either we define a new function to the class to calculate P(X>=x) for upper tail probability, or we can define the cumulativeProbability so that it calculates P(m<=x<=n); in this case if you simply want upper or lower tail, just define m or n separately. and I guess it'll help to short-circuit the class so that it won't produce p>1 or p<0 cases. what do you think? (In reply to comment #2) -- Configure bugmail: http://issues.apache.org/bugzilla/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug, or are watching the assignee. --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
