Title: A distributed algorithm for exhaustive normality test
Authors: Ruben Carvajal-Schiaffino - DMCC - Universidad de Santiago de Chile (Chile) [presenting]
Abstract: The normal distribution assumption of data is necessary for a wide variety of statistical analyses, which are valid once normality tests are passed. Regarding the multivariate normality, algorithms have been developed to test it. However, they are often -near always- applied only to complete samples, excluding subsets of variables. As this number increases, an exhaustive analysis becomes less practical and also a deviation of normality within subspaces in the sample becomes more probable. Most statistical packages run in only one process, without the direct possibility of running distributed algorithms, which could allow a considerable time-saving. A distributed algorithm for running distributed normality tests is introduced. In contrast with the sequential solution, it means a viable solution for this problem.