A Systematic Procedure for Assigning Uncertainties to Data Evaluations

Younes, W

关键词: ALGORITHMS;    BENCHMARKS;    EVALUATION;    PERFORMANCE;   

DOI  :  10.2172/902296
RP-ID  :  UCRL-TR-228283
PID  :  OSTI ID: 902296
Others  :  TRN: US0702922

学科分类：核物理和高能物理

美国|英语

来源: SciTech Connect

【 摘 要 】

In this report, an algorithm that automatically constructs an uncertainty band around any evaluation curve is described. Given an evaluation curve and a corresponding set of experimental data points with x and y error bars, the algorithm expands a symmetric region around the evaluation curve until 68.3% of a set of points, randomly sampled from the experimental data, fall within the region. For a given evaluation curve, the region expanded in this way represents, by definition, a one-standard-deviation interval about the evaluation that accounts for the experimental data. The algorithm is tested against several benchmarks, and is shown to be well-behaved, even when there are large gaps in the available experimental data. The performance of the algorithm is assessed quantitatively using the tools of statistical-inference theory.