Comments on: Rounding Error http://peoplesoftsqr.com/index.php/2010/03/rounding-error/ When Peoplebooks Is Not Enough Mon, 23 Jan 2012 21:54:00 -0700 http://wordpress.org/?v=2.8.4 hourly 1 By: administrator http://peoplesoftsqr.com/index.php/2010/03/rounding-error/comment-page-1/#comment-5892 administrator Wed, 24 Mar 2010 21:01:45 +0000 http://peoplesoftsqr.com/?p=672#comment-5892 Absolutely. I didn't round until right before I wrote the data to the export file, and the export file format imposed the rounding requirements (dollars and cents without a decimal point). The ends-in-fives bias was due to the nature of the input values and type of operations performed, and would have appeared even with infinite-precision arithmetic. Perhaps its similar to the oft-noted bias for numbers to begin with 1. Interestingly, when I first inherited the program, I discovered that some numbers were being prematurely rounded to pennies and others weren't, leading to grand total errors between 10 and 100 dollars. Absolutely. I didn’t round until right before I wrote the data to the export file, and the export file format imposed the rounding requirements (dollars and cents without a decimal point). The ends-in-fives bias was due to the nature of the input values and type of operations performed, and would have appeared even with infinite-precision arithmetic. Perhaps its similar to the oft-noted bias for numbers to begin with 1.

Interestingly, when I first inherited the program, I discovered that some numbers were being prematurely rounded to pennies and others weren’t, leading to grand total errors between 10 and 100 dollars.

]]> By: david nicol http://peoplesoftsqr.com/index.php/2010/03/rounding-error/comment-page-1/#comment-5890 david nicol Wed, 24 Mar 2010 15:22:10 +0000 http://peoplesoftsqr.com/?p=672#comment-5890 presumably you adopted the chemist's best practice of making all calculations using all precision available and deferred rounding to the appropriate accuracy to the display (and comparison) phase? Or were you doing that all along, and wrote this because you had so many ends-in-fives that you multiplied together that the errors got up to the significant part? Premature rounding is a form of premature optimization, when classifying evils. presumably you adopted the chemist’s best practice of making all calculations using all precision available and deferred rounding to the appropriate accuracy to the display (and comparison) phase? Or were you doing that all along, and wrote this because you had so many ends-in-fives that you multiplied together that the errors got up to the significant part?

Premature rounding is a form of premature optimization, when classifying evils.

]]> By: Bob Josephson http://peoplesoftsqr.com/index.php/2010/03/rounding-error/comment-page-1/#comment-5880 Bob Josephson Thu, 18 Mar 2010 02:06:52 +0000 http://peoplesoftsqr.com/?p=672#comment-5880 How about bankers' rounding? See http://en.wikipedia.org/wiki/Bankers_rounding#Round_half_to_even How about bankers’ rounding? See http://en.wikipedia.org/wiki/Bankers_rounding#Round_half_to_even

]]>