A Unified Information-Value Approach to Sensitivity Analysis

敏感性分析的统一信息价值方法

• 批准号: 9515161
• 负责人: Gordon Hazen
• 金额: $ 19.56万
• 依托单位: Northwestern University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1996
• 资助国家: 美国
• 起止时间: 1996-04-01 至 2000-03-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9515161&HistoricalAwards=false
• 关键词: Unified; Information; Value; Approach; Sensitivity

A Unified Information-Value Approach to Sensitivity Analysis Models of phenomena in physical and social science are quite often subject to sensitivity analyses in which analysts evaluate the effects (sensitivity) of the model to changes in parameter values. As models become complex, however, such evaluations becomes cumbersome, time consuming and, at times, suspect õ have all the relevant combinations of parameter values been considered? This research will examine the possibility of using the `expected value of perfect informationn on a parameter or group of parameters as a measure of problem sensitivity to those parameters õ a procedure which is both axiomatically sound and consistent with expected value maximization. A retrospective analysis of standard sensitivity analyses will also be conducted to see how conclusions reached using such methods compare with those suggested by the `expected value of perfect informationn approach.

物理科学和社会科学现象的模型经常需要进行敏感性分析，在这种分析中，分析人员评估模型对参数值变化的影响（敏感性）。然而，随着模型变得复杂，这样的评估变得繁琐、耗时，而且有时令人怀疑õ是否考虑了所有相关的参数值组合？本研究将检验使用参数或参数组上的“完美信息的期望值”作为对这些参数的问题敏感性的度量的可能性õ这一过程在公理上是合理的，并且与期望值最大化相一致。还将对标准敏感性分析进行回顾性分析，以比较使用这些方法得出的结论与“完美信息期望值”方法得出的结论。