Math Puzzle!

OK kids, here's the deal. An economist who shall remain nameless (but he writes for the NYT) thought he blew up the theory that the recession was due to a misallocation of labor across sectors, because the change in unemployment from Dec-07 to Dec-08 was high in several states that didn't experience a big housing bubble.

I thought that was an odd measure, and so I ranked states (A) by their change in unemployment rates from Jun 06 to Dec 08, and then I ranked the states again by (B) their percentage change in house prices from 2q 06 to 4q 08.

I found that 5 of the top 6 states in column (A) were in the top of column (B). So that seems like a very strong relation.

Can someone talk me through how I would compute the chance that this just happened randomly?

For extra credit, how do we deal with the fact that after I saw the results, I came up with the best way to describe them? In other words, I originally looked at the top 10 states in both lists. But since there were no more matches in the 7th through 10th slots, I just dropped them and reported "5 out of the top 6 match." So I'm thinking this cherry picking after the fact may be contributing to the impressiveness of my result, but maybe not?

Comments

  1. I am leaving a comment here so that my email will be tied to this post. I wait with shrinking hope that Wabulon will show me the way.

    ReplyDelete
  2. Anonymous1:23 PM

    Odds that #1 is in one of top 6 spots: 6/50
    Odds that #2 is in one of remaining 5 top spots: 5/49
    Odds that #3 is in one of remaining 4 top spots: 4/48
    Odds that #4 is in one of remaining 3 top spots: 3/47
    Odds that #5 is in one of remaining 2 top spots: 2/46

    100% x (6/50)(5/49)(4/48)(3/47)(2/46)=0.0002832% (1 in 353,127).

    Generalization: odds that the top A of C values are in the top B of C spots is:

    (B! / (B-A)!) / (C! / (C-A)!)

    So the odds that the top 5 were in the top 10 spots is:

    (10! / 5! ) / (50! / 45!) = 0.0119% (1 in 8408).

    Math checks:
    Odds that the top 1 is in the top spot is (1!/0!) / (50!/49!) = 1/50, as expected.
    Odds that the top 50 are all in the top 50 spots is (50!/0!)/(50!/0!) = 1, as expected.
    (note: 0! is defined to be equal to 1.)

    I'm not promising this is right, but it looks good to me.

    Wabulon?

    ReplyDelete
  3. Anonymous2:26 AM

    There was a New wow gold York magazine piece once wow gold about the power struggle
    buy wow gold involved in business lunches.
    buy wow gold It intimated that you
    cheap wow gold could always tell the
    cheap wow gold powerless and the powerful.
    world of warcrft gold The Indians were waiting,
    world of warcrft gold while the Chiefs arrived
    wow power leveling half an hour or
    power leveling an hour later. If
    world of warcraft power leveling you are an On-
    timer, world of warcraft power leveling you
    cannot make an entrance.

    ReplyDelete

Post a Comment

Popular posts from this blog

Libertarians, My Libertarians!

"Machine Learning"

"Pre-Galilean" Foolishness