A comparison of multivariate statistical methods for estimating expected consequences for low‐probability and high‐consequence incidents

Abstract

Abstract The catastrophic incidents involving hazardous materials (hazmats) have often been termed as low probability and high consequence (LPHC). The purpose of this article is to address some fundamental questions with regard to hazmats incidents: What is the expected consequence of a hazmats incident? How should the consequences of incidents involving hazmats be predicted? An exhaustive statistical analysis is performed on the hazmat incident data available from the U.S. Hazardous Materials Incident Reporting System (HMIRS). We present a sequence of logically deduced, linear statistical models to estimate the two major areas of impact of an incident: (1) population affected and (2) cost incurred due to an incident based on the outcomes of the incident. Our initial experiments indicated that linear models are not sufficient for predicting the consequences. Subsequently, we extended our work to evaluate the effectiveness of three multivariate statistical methods, namely (1) partial least squares, (2) spline regression, and (3) Box‐Cox transformations. Based on our experiments, Box‐Cox transformation showed significant improvement in estimating the consequences. Last, we summarize our findings and provide some general guidelines to entities interested in estimating events categorized as LPHC. © 2010 Wiley Periodicals, Inc.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 5 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.