(9c) Statistical Concepts and Data Requirements for Valid Failure Rate Estimation

This paper provides additional details about the statistical concepts discussed in the panel session and the way in which these concepts impact the valid aggregation of field failure data (FFD) used to estimate failure rates. The distinction between point estimation and confidence interval estimation for failure rates is described. Why the upper bounds of failure rate confidence intervals are prescribed by safety standards for use in safety metric computations is explained. Statistical formulas for the failure rate upper bounds are provided. The requirements that the FFD must meet to ensure the validity of the failure rate estimations are discussed from both mathematical and engineering perspectives. The need to understand underlying causes of failure in determining when FFD can be aggregated are explained. Examples illustrating the impacts on failure rate estimation when the data requirements are and are not met are presented.