I can't find any confirmation of my rule, but it's what I was taught. If you always round 5s up, you introduce a bias that is eliminated if you round to the even number, because half the time its up and half the time its down.
Aha, found a reference. Microsoft calls it Banker's Rounding:
Banker's rounding rounds .5 up sometimes and down sometimes. The convention is to round to the nearest even number, so that both 1.5 and 2.5 round to 2, and 3.5 and 4.5 both round to 4. Banker's rounding is symmetric.
This site agrees with -t.
I guess the rational behind this is not to round always up if it's in the exact middle, meaninf 5. Rounding always up will eliminate the "half" nature of that 5, so you have to divide to cases in which you round up and cases in which you round down. I have no idea why the evens get the special treatment, though.
[Edit: x-post wit -t]
Banker's rounding hurts me. In which arenas is it applied?
If you always round 5s up, you introduce a bias that is eliminated if you round to the even number, because half the time its up and half the time its down.
OK, this seems like a good rationale.
Banker's rounding hurts me. In which arenas is it applied?
Well, banking for one. That makes sense, really -- especially if you're dealing with .05's of a cent and whatnot.
This is why rounding numbers is evil. Better just to
live with the fractions!!1!
Well, banking for one.
Get the fuck out of town. Any others? I mean, is there a reason for it to be applicable to GPAs?
Better just to live with the fractions!!1!
Tell that to 3.14159265.
This is why rounding numbers is evil.
I don't think rounding is always evil and the irrational numbers support me in e-mail.
I mean, is there a reason for it to be applicable to GPAs?
For grading on a curve? Yeah, I'm just making shit up now.
Any others? I mean, is there a reason for it to be applicable to GPAs?
My guess (and it's the first time I've heard of the rule either, so it's only a guess coming out of my sieve) is that it's a rule for statistics, for when you're using lots of numbers, when you don't want to skew an average or change the sum significantly. Since statistically half the numbers that end with 5 end with odd-5 and the other half ends with even-5 that's just a way to make sure you round-up around the same times of numbers that you round-down, so the sum - and the average - aren't changed.
For something like one perosn's grades, though, just the single number, no statistics nvolved, I don't see how it is needed. Unless, of course, you want to average out grades from students all over. It's still not needed for each single student, but for a group, maybe.
