3 ms·
The thing about power laws is that they assert an underlying mechanism that causes the power law, but they look very close to an exponential or other long-taile
by j2kun 3y ago
The thing about power laws is that they assert an underlying mechanism that causes the power law, but they look very close to an exponential or other long-tailed distribution, and it's not so easy to distinguish between them empirically.
I recall reading a paper (but I can't find it right now) that re-evaluated a bunch of datasets that were claimed to exhibit power laws, and found that they actually were more closely aligned to an exponential distribution or other.
- timy2shoes 3y agoWas it Clauset, Shalizi, and Newman 2009? That’s the one I know, but I believe there’s many more, especially from Clauset. https://www.stat.cmu.edu/tr/tr853/tr853.pdf https://www.stat.cmu.edu/tr/tr853/tr853.pdf
- creer 3y agoIn this article, they seem to simply mean "power law-like distributions — meaning a few massive hits and a vast number of misses". ... But then they try and give two mathematical chapters in the presentation! Ugh.