On Skewed Distributions and Straight Lines

In this paper, we present a hypothesis that power laws are found only in datasets sampled from a static data, in which each and every item has gained its maximal importance and is not in the process of changing it during the sampling period. We motivate our hypothesis by examining languages, and word-ranking distribution as it appears in books, and in the Bible. To demonstrate the validity of our hypothesis, we experiment with the Wikipedia edit collaboration network. We find that the dataset fits a skewed distribution. Next, we identify its dynamic part. We then show that when the modified part is removed from the obtained dataset, the remaining static part exhibits a good fit to a power law distribution.