Tuesday, April 16, 2013

Math In The News

You've got to love "practical applications" of the normal curve:
While yesterday's cliff-dive in gold was impressive by any standards, the escalating drop over the past 5 days has been just as dramatic. Based on 20 years of rolling 5-day moves, the ~15% plunge is equivalent to around 7 standard deviations (in context Yao Ming is a mere 6 standard deviations taller than the average human making gold's move the equivalent of meeting a man taller than 7'7")
Too bad the reason behind it is so bad.

1 comment:

Anonymous said...

I would be *more* impressed if the distribution of 5-day gold returns was normal. Or log-normal.

But it isn't.

A 7-sigma event for a non-normal distribution means ... what? Maybe a good question for you stats professor because I *really* don't know.

In any event, there is a medium amount of empirical evidence that stock (and commodity and ...) returns follow a Levy (also called Pareto-Levy) distribution. One interesting feature of this is that the distribution is non finite. The more samples you take, the wider the outliers (I phrased that poorly...)!

-Mark Roulo