Article ID: | iaor20165094 |
Volume: | 13 |
Issue: | 4 |
Start Page Number: | 243 |
End Page Number: | 277 |
Publication Date: | Dec 2016 |
Journal: | Decision Analysis |
Authors: | Keelin Thomas W |
Keywords: | simulation |
The metalog distributions constitute a new system of continuous univariate probability distributions designed for flexibility, simplicity, and ease/speed of use in practice. The system is comprised of unbounded, semibounded, and bounded distributions, each of which offers nearly unlimited shape flexibility compared to previous systems of distributions. Explicit shape‐flexibility comparisons are provided. Unlike other distributions that require nonlinear optimization for parameter estimation, the metalog quantile functions and probability density functions have simple closed‐form expressions that are quantile parameterized linearly by cumulative‐distribution‐function data. Applications in fish biology and hydrology show how metalogs may aid data and distribution research by imposing fewer shape constraints than other commonly used distributions. Applications in decision analysis show how the metalog system can be specified with three assessed quantiles, how it facilities Monte Carlo simulation, and how applying it aided an actual decision that would have been made wrongly based on commonly used discrete methods.