VALIDATION OF PROBABILISTIC MODELS
作者:
NIELSC. LIND,
期刊:
Civil Engineering Systems
(Taylor Available online 1996)
卷期:
Volume 13,
issue 3
页码: 175-183
ISSN:0263-0257
年代: 1996
DOI:10.1080/02630259608970196
出版商: Taylor & Francis Group
关键词: Entropy;forecasting;information;generator;likelihood;models;probability;seismic;earthquake;validation;verification
数据来源: Taylor
摘要:
Probability statements in applied science cannot be verified, i.e., shown to be true. But, as shown, the models or methods that generate probability statements can be validated in the sense that they can be assessed using real or simulated data. The performance of probabilistic models can be compared in terms of likelihood, Shannon information or entropy. Likelihood and information provide equivalent measures of relative validity. Examples drawn from weather forecasting and seismic engineering illustrate the approach.
点击下载:
PDF (304KB)
返 回