I found this article when searching for information on statistical methods based on power laws. I've been trying to find information on non-Gaussian statistics, but having a difficult time making my way by myself. I even bought a book of non-parametric statistical methods, but even those ended up using z-scores based on a normal distribution when dealing with large samples.
While the blog post is a bit airy, I appreciate its point that the social sciences in particular are more accurately described by power laws than by the bell curve.
Yeah, this was a really interesting read. In my experience, power-law distributions seem to more accurately describe more of the world than bell-curves. Yet we seem to have better tools for working with bell curves (at least in my highly limited experience working with statistics).
Nassim Nicholas Taleb's books (Fooled by Randomness and The Black Swan) make the point that power laws are better at describing how uncertain the world we live in is, rather than making uncertainty and risk tractable in the way that the bell curve models.
The properties of the bell curve are well known, and therefore easier to perform statistical tests with, whereas the many stable power law distributions are harder to deal with and more difficult to simply assume.
I'm curious, on a biological level, if this is one of the assumptions our sensory and cognitive systems make to filter out lots of noise and find patterns.
4 comments
[ 5.2 ms ] story [ 19.8 ms ] threadWhile the blog post is a bit airy, I appreciate its point that the social sciences in particular are more accurately described by power laws than by the bell curve.
The properties of the bell curve are well known, and therefore easier to perform statistical tests with, whereas the many stable power law distributions are harder to deal with and more difficult to simply assume.