Visual Interactive Fitting of Beta Distributions

Abstract

This paper describes a visual interactive procedure for fitting beta distributions to activity times in a simulation model when sample data are not available for statistical analysis of the model's input processes. Using subjective information about a given activity time the modeler specifies the activity's minimum and maximum times together with two of the following characteristics: mode, mean, variance, or selected percentiles. The fitting procedure includes efficient methods for computing the shape parameters of the beta distribution that most nearly matches the specified characteristics. The user can manipulate the fitted distribution by either revising the specified characteristics or directly altering a visual display of the density. This fitting procedure was implemented in a public‐domain microcomputer‐based software system called VIBES (visual interactive beta estimation system). An example from construction engineering illustrates the operation of VIBES.