A framework estimating the minimum sample size and margin of error for maritime quantitative risk analysis
本文提出了一个系统框架，用于从事故数据库中估算平均事故频率所需的最小样本量和误差范围，并应用于海上事故数据集，发现现有数据库样本量不足，需考虑至少10%的误差范围。
The average accident frequency is essential for quantitative risk analysis and is conventionally estimated from accident statistics. This paper has systematically synthesised the knowledge on statistical errors and offered the missing instructions, a framework, for determining the minimum sample size and the margin of error (MOE) when calculating the average accident frequency from an accident database at hand. We have applied this framework to representative accident datasets in the maritime domain and presented the revealing results that can already be used in QRAs based on these datasets. The findings are useful to both QRA analysts and policy makers. Interestingly, the framework application has revealed that the determined minimum sample sizes would exceed the datasets available in existing maritime casualty databases by decades, requiring at least 10% MOE to be factored into pertinent QRAs. By the same token, the earlier notable QRAs (developed as part of formal safety assessments in support of rule making) had to consider the MOE of over 30%, given the sample sizes used, likely shifting the conclusions they arrived at. Other findings of the application have shown that the average accident frequencies for large passenger ships have remained constant over the past 40 years.