Recently, I read papers that perform power-law fitting on their empirical data (estimate the alpha), some of them report corresponding p-value for the Kolmogorov-Smirnov test, but many of them do not.

I am completely new to this kind of work and I am able to perform power-law fitting thanks to the program from Clauset et al. However, according to the K-S test and its corresponding p-value, my data seems doesn't follow a pure power-law distribution.

Then I use the python program powerlaw to do a truncated power-law fit, which seems a better fit visually, but this package doesn't provide a K-S test function.

And my question is:

  1. How to perform a Kolmogorov-Smirnov test on the truncated power-law distributed data? or maybe other tests?
  2. How to estimate the xmax of a truncated power-law?
  3. Is it normal that a power-law fitted line does not go through the data points, but lies above them?

I think this is a very hard problem for me, is there some concrete tutorial on data fitting and testing?

Here is my data and the power-law fitted line (alpha=1.92) enter image description here

⬇ My data and the truncated power-law fitted line (alpha=1.66, beta=0.0009) enter image description here

⬇ The truncated power-law fit of the data from a paper enter image description here


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.