Considerations When Simulating Data From the g-and-h Family of Distributions

• In Event: Thursday Roundtable Session 9:00 am
  In Roundtable Session: AI, Psychometrics, and Data Analysis (Table 24)

Abstract

The g-and-h family of distributions is a popular, efficient, and flexible option to model and simulate non-normal data. In spite of its popularity, there are several aspects of these distributions that need special consideration when they are used. In this paper three will be presented: (a) there exist more than one combination of (g,h) parameters that can generate the same population univariate skewness and kurtosis; (b) only one of the two available multivariate generalizations in the literature is a true g-and-h distribution; and (c) the multivariate generalizations available are actually special cases of the same family of multivariate distributions, the Gaussian copula. A small simulation showcasing the method of homotopy continuation to obtain (g,h) parameters is also introduced.

Author

• Oscar L. Olvera Astivia, University of Washington