Monday, March 25, 2013

Monday: There was going to be math here

Unfortunately, I'm not sure that the intended math was actually the problem with things, so I'm going to skip that for now.  I now have some concern that the eta parameter I've been assuming is defined only on [0:1] probably isn't actually bounded on the large side.  The approximation I'm using needs to be, as the expansion around zero needs higher order terms, and I don't feel like deriving them.

In any case, I don't think that actually solves the main issue I'm seeing, which I think is a generic failure of the model to match observations as eta->infinity.

On top of all that math crap, I apparently can no longer grill a cheese.  My last two attempts have had side A ending up burnt, despite me doing things exactly the same as I've always done.  Maybe my stove has recalibrated itself hot?

Ok, I'm writing this post, and I just saw that MSNBC is having a story about a fucking groundhog not predicting the weather correctly.  An animal.  Didn't predict the weather correctly.  You can take a moment if you want.


Then, just to make my brain try to burn down everything, they posted this graphic:
Thanks for making the video available online.
Ok, so 39% over 127 years, and this was claimed as support for an animal not being "that bad" at predicting weather.  That works out to 44 successes in 127 trials, which just happens to be why I have a script named "even_coin.pl" that calculates the Bayesian PDF for a given coin.  This is conveniently just the binomial theorem, such that:
PDF(H) = (n!/(h! (n-h)!)) H^h (1-H)^(n-h)
Plotted up for this case.
The assumption would be that an even coin is distributed around 0.5.  As you increase the number of trials, the width of the distribution shrinks.  In any case, I'm not entirely sure that you'd use this distribution to claim that a groundhog is better than randomly flipping a coin.

"Why don't we teach math better?"

  • I haven't actually read through this too much.  My quick understanding is "better than previous plans," but "still pretty bad."  One odd thing I heard on the radio was that the bank holiday is going to extend until Thursday, as there's concern that if they opened on Tuesday, there'd still be a run on them.  If I'm going to panic withdraw all my money on Tuesday, I'm going to be more panicky withdrawing it on Thursday.
  • I wonder if The Bloggess has heard of this.  It's in NYC, though.  Also: "Please do not bring any dead animals with you to the class."
  • Squirrel!
  • Imagine how weird it would be if modern sports bars required this level of attire.

No comments:

Post a Comment