Used to estimate the likelihood of each undesired event identified at the hazard identification stage. The purpose of frequency analysis is to determine the frequency of each of the undesired events or accident scenarios identified at the hazard identification stage. Three basic approaches are commonly taken: (1) use relevant historical data to determine the frequency with which thexe events have occurred un the past and hence make judgements as to the frequency of their occurrence in the future. The data used should be relevant to the type of system, facility or activity being considered and also to the operational standards of the organization involved; (2) predict event frequencies using techniques such as fault tree analysis and event tree analysis. When historical data are unavailabre or inadequate, it is necessary to derive event frequencies by analysis of the system and its associated fault modes. Numerical data on all relevant events, including equipment failure and human error from operational experience or published data sources are then combiend to produce an estimate of the frequency of the undesired events. When using predictive techniques, it is important to ensure that due allowance has been made in the analysis for the possibility of common mode failures involving the co-incidental failure of a number of different parts or components within the system. Simulation techniques may be required to generate frequencies of equipment and structural failures due to ageing and other degradation processes, by calculating the effects of uncertainties; (3) use expert judgement. There are a number of formal methods for eliciting expert judgement which make the use of judgements visible and explicit and provide an aid to the asking of appropriate questions. Expert judgements shuold draw upon all relevant available information including historical, system-specific, experimental, design, etc. The methods available include the Delphi approach, paired comparisons, category rating and absolute probability judgements