Critical truths about power laws

From Brede Wiki

Jump to: navigation, search
Paper (help)
Critical truths about power laws
Authors: Michael P. H. Stumpf, Mason A. Porter
Citation: Science 225 (6069): 665-666. 2012 February
Database(s): Define pmid
DOI: 10.1126/science.1216142.
Link(s):
Search
Web: Bing Google Yahoo!Google PDF
Article: BASE Google Scholar PubMed
Restricted: DTU Digital Library
Other:
Services
Format: BibTeX
Extract:

Critical truths about power laws is a commentary that critize applications of power laws when interpreting data.

 y(x) \approx x^{-\lambda} \quad \text{for} \quad x > x_0

As a rule-of-thumb they regard a candidant power law should exhibit "an approximately linear relationship on a log-log plot over at least two orders of magnitude in both the x and y axes".

They give examples where power laws have been reported:

  1. Gutenberg-Richter law in seismology
  2. Allometric scaling in animals
  3. World Wide Web hyperlinks
  4. Urban living
  5. Insurgent and terrorist activity
  6. Scientists publication rates

They claim good well-supported power laws in:

  1. Magnetization
  2. Allometric scaling in animals

while arguing that numerous scholar have failed to apply statistical test to their data just "eyeballing" straight lines. Furthermore, they argue that the arguments on power laws might lack a generative model.

They give examples were argued power law does not hold (in their opinion):

  1. yeast protein[1] citing an critical analysis by Aaron Clauset.[2]
  2. C. elegans nervous system

Of relevance Aaron Clauset, Joel Ornstein, Adam Ginsburg, Wim Otte and Laurent Dubroca have produced some statistical software to handle power laws:[3]

http://tuvalu.santafe.edu/~aaronc/powerlaws/

[edit] References

  1. High-quality binary protein interaction map of the yeast interactome network
  2. http://www.cs.unm.edu/~aaron/blog/archives/2008/10/power_laws_in_t_1.htm
  3. Power-law distributions in empirical data
Personal tools