Picture of virus under microscope

Research under the microscope...

The Strathprints institutional repository is a digital archive of University of Strathclyde research outputs.

Strathprints serves world leading Open Access research by the University of Strathclyde, including research by the Strathclyde Institute of Pharmacy and Biomedical Sciences (SIPBS), where research centres such as the Industrial Biotechnology Innovation Centre (IBioIC), the Cancer Research UK Formulation Unit, SeaBioTech and the Centre for Biophotonics are based.

Explore SIPBS research

A comparison of data-driven and model-based approaches to quantifying railway risk

Bedford, T.J. and Quigley, J.L. and French, S. (2004) A comparison of data-driven and model-based approaches to quantifying railway risk. In: Probabilistic Safety Assessment and Management. Springer-Verlag, London, pp. 2765-2771. ISBN 9781852338275

[img]
Preview
PDF (BedfordQuigleyFrenchPSAM2004-approaches-to-quantifying-railway-risk)
A_Comparison_of_Data_Driven_and_Model_Based_Approaches_to_Quantifying_Railway_Risk.pdf - Preprint

Download (127kB) | Preview

Abstract

This paper presents some of the results of a project sponsored by the UK Railway Safety and Standards Board (RSSB). An earlier statistical evaluation of a previous version of the RSSB Safety Risk Model (SRM), a combined Fault/Event Tree, conducted by Prof Andrew Evans had concluded that the model was unduly pessimistic. We have constructed a hypothesis test based on the relative likelihood techniques using the most recent version of the SRM as the null hypothesis. The results support the SRM being consistent with the historical data. Two significant differences between these two studies are the statistical methods employed to support the analysis and the removal of certain significant conservative assumptions from updating the versions of the SRM. The paper discusses the demands that different model purposes place on these models, and explores the question of whether or not it is meaningful to compare their outputs. The use of expected fatalities as a metric for expressing risk in both models is questioned because of the heavy-tailed form of the distribution for fatality numbers given a fatal accident.