78.4%

Some thoughts on Nate Silver’s latest prediction:

…one can think that Silver is probably right about the Electoral College, and simultaneously think that the 78.4 percent number is basically meaningless. Or rather, that it is impossible to formulate the epistemological difference between “There is a 78.4 percent chance that Obama will win” and “There is a pretty good chance Obama will win.”

My problem with it is that I don’t believe that he knows all of his data inputs to three figures. Yes, I know that’s how the polls purport to measure them, but three figures of precision are meaningless unless you also believe that the number is accurate. As another plug for my (still to be published) space safety book, here’s a relevant excerpt:

One of the very first things that scientists and engineers are taught is that you can’t get an answer more precise than the precision of the least precise factor from which it is derived. For example, we know the gravitational constant to many places, but when you multiply it by a mass that you only know to two places, that is the maximum precision that can be reasonably used to express the local gravitational field for that body. A good professor will mark down an answer on a student’s test that, while accurate, is unjustifiably precise. When I see engineers doing the same thing, I tend to think that they’re trying to impress the innumerate who don’t understand the difference between precision and accuracy. And I think that the safety numbers for Ares I were precisely wrong.

As is Nate’s election prediction. To me, it would be more credible if he just said 80%, though I still don’t buy it.

14 thoughts on “78.4%”

  1. Perhaps not the best example. Mass of an astronomical body, if it’s not just a guess, is usually calculated backwards from the gravitational field and the gravitational constant.

    1. Yeah, Daver. According to the multiworlds interpretation of quantum theory, you’ve just created a new universe where kids do science right and grow up to be mature adults. Why couldn’t someone have done that for our universe?

      1. I don’t suppose that world has a lot of cute moms willing to express their appreciation in a fairly primitive fashion.

  2. They gotta try everything possible to make the voters ‘think’ it’s over. Just maybe, they can fool enough people will stay home, so that it will make a difference, and Obumble will win.

    Personally I think this will be Carter vs Reagan all over again. The talking heads keep saying it’s close, it’s a toss up, its so close we don’t know, blah, blah, blah. But I don’t know anyone who voted for Obumble before who is doing so again. And I’ve got some hyper-lib friends. But they’re mostly business owners too, two of whom told me it’s a choice of lifelong ideology vs the life of their businesses.

    I just didn’t bother to say I told you so. One of them was mad at me just few weeks ago about an Obama joke I sent out. [it was an accident…no, seriously!] So I did ask her WHY she got so mad, but she’s not voting for him.

    “…what choices do I have? Romney is a business guy, I hate voting for him, but there’s no such thing as a smart, Democrat businessman to vote for anymore!”

    I can’t remember a smart business guy who ever ran for the WH anyway. But she was already already just about to weep, SOOOooo I let it go.

    1. I used to work for a couple ultra-liberals in a mom and pop company. They both voted for Ralph Nader at the time. One day they were wondering why their business was doing well under Clinton when they recognized he’d made some bone head decisions. I told them, “you built this business, not the government” (was I being prescient or what?) “You and others like you are why Clinton is doing so well.”

  3. Interesting take from Breitbart….

    http://www.breitbart.com/Big-Government/2012/11/01/Eight-reasons-Pro-Obama-Polls-Are-Wrong

    Point #8 was key for me, where they suggest that one should look to Gallup and Rasmussen polls for the greatest accuracy. That’s all they do for a living and they have to be accurate. The other polling companies are tied to universities and media companies who are swarming with liberal agendas.

    Romney/Ryan are looking like they will win by 7 points or greater.

  4. Silver runs thousands of simulated elections through his model. “Obama has a 78.4% chance of winning the Electoral College” just means “78.4% of the simulation runs have Obama winning the Electoral College.”

      1. I am not a mathematician, but I don’t think it’s that simple. Say you measure the height of the Army’s recruits every year, with a precision of 1 inch. One year the average height is 70.4″, and the next year it’s 70.5″. The difference is to a higher level of precision than the original measurements, but it’s still legitimate to say that the average recruit height has increased, because you’ve averaged thousands of those measurements, and there’s no reason to expect the net measurement error to favor one year over another.

  5. Silver was also off by 20% in his perdiction of the 2010 Mid-Term…GUESS WHICH DIRECTION HE WAS OFF IN!!!!!!!

  6. I believe I learned about significant digits all the way back in high school (>40 years ago), and having a degree in computer science makes me all the more wary of trying to derive more precision from calculaations or models than you should. (And, I’ll add, my graduate class in numerical methods, combined with my own digging into climate change models and data, is why I’m a profound AGW skeptic.) I suspect Silver’s “prediction” is largely driven by his assumptions and choices (as I think Jim Gerghty demonstated this morning).

Comments are closed.