When Errorbars Hit Mainstream News
It’s interesting to me how science has come into conflict with the news in the past week. First we had the deflategate, where footballs mysteriously deflated during a playoff game, and then we had an over hyped blizzard.
The NFL recently hired physicists at Columbia to help make the case for science with the football fiasco, but I think that’s unnecessary: a few good experiments with temperature and friction and lots of measurements by lots of different pressure gauges will empirically demonstrate how much of a range we might expect from such things. In other words, understanding errorbars.
As for the blizzard, this article nicely articulates the science of weather forecasting and what went wrong. But what is interesting is that, in general, models have gotten much better, and in particular are good at predicting how powerful a storm is going to get. In this case the model got that right, but then the error came in figuring out exactly where the storm would travel and when.
Again, it’s a case of errorbars, and the public seems not to understand it. Or maybe they just don’t want to.
In fact, I heard quite a few people call in to ESPN radio over the past week trying to explain to the sports radio hosts what might be going on scientifically, only to be hung up on. The truth is, it’s not as interesting a story to think about it just happening outside our control. It messes with our sense of omnipotence and control.
This is bad news for society, as more and more things become “datafied” and as we assume that will translate into perfect information.
mathbabe 
Maybe this is too fine a point, but I think the causation runs the other way:
we have a strong bias to think that historical events were certain to happen. One consequence is that winners don’t sufficiently credit being lucky and losers don’t sufficiently credit being unlucky. Also, it has led people to think that if we only had enough favour with the gods/favour with The God/wisdom/data+computing power/etc then we could really know the true outcome in advance.
LikeLike
Here’s another place where people don’t seem to know the range of error. I am a big fan of slower traffic keep right, and keep right to pass laws that exist in many states. I hear some people saying, in defense of driving in the inside lane(s), I was going the speed limit, why can’t I drive where I want?
So practically, the bigger issue, is that your speedometer doesn’t show you an accurate speed in most cases. Like in deflate gate, the diameter of your tires is the controlling factor. Most speedometers use the RPM of the transmission output shaft, or some other “rotating” rod and a “counter” to then do some math, either mechanically (in most cases), or electronically (in newer vehicles), using the diameter of your tires, to tell you how fast your vehicle is “going”.
As your tires wear, or if the inflation is not exactly right, that speed is usually going to be between 1.5 (typical sized tires) and as much as 3.5 mph slower than what your gauge says. Under-inflated, larger diameter, worn tires like on a semi tractor where the radius change of 1/8 to 3/16 of an inch contributes to a larger circumference change, may be more.
My experience on the road, over the years, has taught me that most people are pretty clueless to traffic management, on their own part. Alas, congestion is caused by factors we all know as queuing theory basics. The goes-ins must be less than or equal to the goes-outs, or things start backing up. That’s simple math. As well, in rush hour traffic, the reason you quickly accelerate to speed as you leave the light, is not so that you can get to the next stopping point faster. It is, in fact, so that the people behind you can also make it through the light so that the goes-outs can be as high as possible to accommodate all the goes-ins behind you. Still, simple math that people don’t always do, because most are distracted with talking on their phones, primping hair, or conversations with their passengers.
LikeLike
Wow, I’d never heard this speedometer/tire-diameter connection before (I just never thought about it before). Do you know a good source (URL) on the Web for a fuller discussion of it? (might make for an interesting blogpost).
LikeLike
The simplest explanation is that you just have to know geometry and understand that the circumference of your tire in inches, times the RPM gives you inches per minute. Multiply by 60 (for inches per hour) and divide by 12 (for feet per hour) and then by 5280 (for miles per hour). Now, if you change any of those factors, and in particular the diameter of your tire, you will arrive at a new number.
LikeLike
Thanks, I get that, I think I just always naively assumed the speedometer was taking into account some other factors as well… anyway, next (naive) question then: is the odometer also just based on the tire diameter, such that over a long period of time it will always read higher than the true mileage?
LikeLike
IMHO, even if you are going the speed limit and your tires are large enough to give you an accurate reading, it’s extraordinarily obnoxious to hang out in the left lane while there are people going faster than you trying to pass. In fact, it’s illegal in most states.
http://www.vox.com/2014/6/16/5804590/why-you-shouldnt-drive-slowly-in-the-left-lane
LikeLike
Errorbars are completely overlooked in the reporting of standardized test scores where the public is fed seemingly precise rankings of schools based on test scores that have varying margins of error that depend on the size of the school… and I know you are wise to the VAM scam which ignores error bars altogether…
LikeLike
Probably the most error bar is the one associated with counting the votes in the disputed Florida election of November 2000.
LikeLike
Arg… should have read “most famous error bar”
LikeLike
Nice blog. I was trying to use the code in this quite old post:
(where the comments are closed)
Is that code absolutely correct? I was wondering about a line in the function get_vol(d):
var = (1-1.0/lam)*var + (1.0/lam)*r**2
shouldn’t it be:
var = (1-1.0/d)*var + (1.0/lam)*r**2
so that we can follow the formula
$V_{new} = s \cdot V_{old} + (1-s) \cdot r_0^2$
ie, shouldn’t lam be d in (1-1.0/lam)?
If I understood correctly, lam is just the sum of decays for a given number of returns being used in the volatility calculation, thats why we divide the new return by lam. But the old var should be multiplied by s, which is 1-1.0/d, right? So why isn’t it, or am I making a mistake??
Would appreciate help, though I can see this is an old post!
LikeLike
No, I don’t think so. The weights should add to 1.
LikeLike
I think the real bad news for society is that this “scandal” represents the most frequent use of the term deflation in the last six years. There are so many dimensions to Operation Dumbdown it is hard to keep track.
LikeLike