Because if you're attempting to fit a polynomial to something that spans 4 orders of magnitude and has outliers that can cover that entire range, you're going to have problems. It turns out that one point at 1e3 is equivalent to 100 points at 1e1. However, taking logarithms of things shrinks everything down so that they have roughly equivalent weights on the fit. I'm still not totally happy with the way this transforms the error estimates, but it at least gives a better polynomial fit. You know, one that actually looks like the data.
 |
| "My hat has a hat. It's a fox." |
- I've heard a lot of people wondering why people are glad that Thatcher's dead, even if they were too young to live under her government. I suspect things like this are the reason. She was kind of a horrible person.
- Alphanumeric!
- Yes. I'm sure it was just like this. Dole plantation here on Oahu nicely fudges the history, where they point out all the nice people who stayed in houses on the plantation that the company provided so they wouldn't have to travel so far to work.
- Bunnies.
- I never really did a science fair.
No comments:
Post a Comment