Some Thoughts on Bias Corrections

The problems associated with making bias corrections to SNM accountability data have received considerable attention recently. In particular, two key questions are: (1) when should bias corrections be made, and (2) whether or not bias corrections are made, how may the uncertainty associated with the resulting data be expressed? There is no simple answer to question (1). The problem is complicated by the fact that as standards, procedures, and regulations are developed, the administrative problems associated with making bias corrections cannot be ignored. These problems occur because of time lags in standards data such that when bias corrections are required, this may mean that a substantial body of past data needs revision. A common action is to make such corrections on subsequent data also, and here the assumption is implicitly made that the bias estimated for one period of time will also apply to the next time period. Such an assumption is not necessarily valid because of changing conditions in the laboratory. It is the purpose of this paper to present some discussion and results that are hopefully of value in finding answers to the questions given in the first paragraph.