CMStatistics 2017: Start Registration
View Submission - CMStatistics
B0877
Title: Modelling real phenomena with power law tail by the family of generalized power law distributions Authors:  Faustino Prieto - University of Cantabria (Spain) [presenting]
Jose Maria Sarabia - University of Cantabria (Spain)
Abstract: Power laws (Pareto distributions) are very common in physics and social sciences. However, they are usually valid only in the upper tail of the distribution. We explore the properties of a new family of distributions - the family of Generalized Power Law (GPL) distributions - that we could use to model, in the whole range, real phenomena with power law tail. In addition, we provide empirical evidence of the efficacy of those distributions with real datasets. To do that, we use the following methodologies: (1) maximum likelihood method for fitting the models to the datasets; (2) Bayesian information criterion for comparing them with other well-known models; (3) rank-size plot as a graphical method; (4) Kolmogorov-Smirnov test method based on bootstrap resampling for testing the goodness of fit of those models.