Most of dielectric relaxation data were able to be modeled by the final fitting law: the combined CS + HN laws.
But when the European Commission's draft proposals for digital copyright reform were published last September, they were criticized by tech companies as regressive, and by copyright reformists as a missed opportunity to modernize ill-fitting laws to make them fit for purpose in the internet age.
Thanks also to Aaron Clauset, Cosma R. Shalizi and Laurent Dubroca for posting methods for fitting power laws and generating power-law distributed random variates in R.
Note that the methods of Clauset et al [14] for fitting power laws to empirical data include an estimation of a lower threshold of event duration, below which the distribution does not exhibit power law behavior.
For instance, the χ2 value for a lognormal distribution fit to a community created with a power-law distribution is 65.7, whereas the best fitting power-law estimate for that community gives a χ2 value of 60.1.
This method starts from the real data and obtains the exponent of a best-fitting power-law, α, by maximum likelihood estimation.
Mukherjee et al. experimented with fitting inverse power laws to empirical learning curves to forecast the performance at larger sample sizes [ 1].
The clustering coefficient distribution does not follow a power law, thus, the results of power law fitting of clustering coefficient distribution were: r = 0.120; R2 = 0.232.
The clustering coefficient distribution does not follow a power law, thus the results of power law fitting of clustering coefficient distribution were: capacitation: r = 0.152, R2 = 0.194; AR: r = 0.023, R2 = 0.132.
This suggests some asymmetry of the goodness of fit testing: whereas increased sample size minimizes incorrect power law fitting of a multi-exponential process at sample sizes of n = 640 (Figures 3,6,7), power law data could still be incorrectly fitted with multi-exponential functions in a majority of trials due to the larger number of free parameters (Table 1).
A value of zero (green) means the power law fitting was always rejected, while a value of 1 (red) means that the power law was never rejected.
