Development of a Software Module for Statistical Analysis of Variability and Uncertainty

A software tool, AuvTool, for statistical analysis of variability and uncertainty associated with fitting distributions to data sets is being developed for use with the Stochastic Human Exposure Dose Simulation (SHEDS) modeling framework. It is also generally applicable for quantifying variability and uncertainty in data sets used for risk assessment, emissions estimation and other quantitative analysis fields. The tool features the use of bootstrap simulation and twodimensional Monte Carlo simulation for simultaneously quantifying variability and uncertainty. Object-oriented programming techniques are applied to the development of the software tool. This paper presents the software structure design and development considerations, and introduces the main function modules and features included in the software tool. These features are: 1) an input data sheet similar to a spreadsheet for data input, which also can import or export Microsoft Excel and tab-limited text files; (2) parameter estimation for common parametric distributions; (3) Kolmogorov-Smirnov and Anderson-Darling statistical test measures to help the user choose a good fit; (4) batch analysis to help the user to handle a large number of datasets efficiently; (5) uncertainty analysis for variables without original data; (6) further analysis of the sampling data of statistics of interests from bootstrap simulation; and (7) instant graphical presentation of simulation results and tabular summarization of variability and uncertainty analysis results. In most cases, two parameter estimation methods, including method of matching moments and maximum likelihood estimation, will be available. An example case study is used to illustrate the application of the software tool.