maxbiostat / bayesian_power_law

A Bayesian take on fitting power law (and alternatives) to data

bayesian_power_law

A Bayesian take on fitting power law (and alternatives) to data

bayesian_power_law's Issues

list of papers to cite/have a look

. Zwetsloot, 2018

compare LRT with BF

Sketch:

• a) Simulate 1000 data sets from a power law with $\alpha = 2.5$ and $x_min = 1$, 50, 000 observations.
• b) Simulate 3 X 1000 data sets from a moment-matching log-normal with mean = 3 and cv = {0.1, 1, 10};
• c) Simulate 1000 data sets from a moment-matching exponential with mean = 3;

Analysis

a) For each simulation, pick x_min using CSN then apply the LRT of power-law vs log-normal (LRT_0) and power-law versus exponential ( LRT_1);
b) For each simulation, pick x_min using CSN then apply LRT_0;
c) For each simulation, pick x_min using CSN then apply LRT_1;

Repeat the above analysis for Bayes factors.

Compute the error rate of each "test".