To avoid double counting of empirical data, you must not apply the test or field data collected at a parent level to a child level. For example, if an assembly is tested for 1000 hours and experiences failures, this data should be applied only to this assembly. You should not try to break the data out further or reference it again at the subassembly level. The effect of doing this would be to double count the data, thereby incorrectly biasing the calculated assembly failure rate.

You also do not want to use Bayesian analysis to modify the failure rate for an assembly if field failure data is already being taken into account. For example, a specified failure rate may already include field failure data. Or, field failure data may be specified using an adjustment factor or Telcordia calculation method. In these cases, you do not want to use Bayesian analysis with this assembly because the adjustment for field failures would be accounted for twice.

An assembly can contain parts with specified failure rates and parts with calculated failure rates. The parts with specified failure rates may already include field failure data adjustments. When Bayesian analysis is applied at the assembly level, parts with both specified and calculated failure rates are considered. Thus, for parts with specified failure rates, you could be accounting for the field failure data adjustment twice. For more information, see Failure Rate Adjustments.