Experimental investigation of the beta distribution in Monte Carlo simulation
Date
1981
Authors
Ganeriwala, Manju Surendra
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Department
Description
The primary concern of this report is to statistically analyze and test the possibility of substituting the use of beta distribution in simulation models using Monte Carlo technique, with some other distribution(s) which closely resemble the characteristics of the beta distribution. The two proposed substitutes are the triangular distribution and the lognormal distribution. A computer program with several subroutines is designed for the purpose of experimental investigation. Given the parameters, random samples are drawn from the triangular, the lognormal and the beta distribution. Later the triangular and the lognormal sample is compared with the beta sample for testing the goodness of fit at a given level of significance. A number of computer runs are made each time with a different range, different set of parameters and different sample size. Another problem addressed in this report is the computation time required for sampling from beta distribution. In general, random sampling from beta distribution, irrespective of the form employed (the actual or the simplified version) consumes much more computer time as compared to sampling from the triangular or the lognormal distribution, thus increasing the cost of simulation