R Code for Data Simulation with Moment Matching

Simulation from data may be carried out for various purposes in water resources and is most simply achieved when the recorded data set can be regarded as a sequence of independent random variables. The aim is then to simulate a data set from a proabability distribution with similar statistical properties to the recorded data. In a recent paper it was shown that a finite mixture of a large number of generalised beta distributions can be utilised to simulate univariate data which has the same mean, variance, skewness, and kurtosis as any recorded data (Bardsley, 2017). Attention is drawn here to an R implementation of the simulation procedure of that paper. The code is reasonably efficent and generating a million data values requires about one minute on a standard PC.