Limit of Error - Who Needs It?

The regulations of the Nuclear Regulatory Commission (NRC) and the orders of the Department of Energy (DOE) require that a l i m i t of error (LE) be calculated for inventory differences (ID), for shipper/receiver differences, for transfer of nuclear materials and for almost any other item requiring a measurement. With such important governmental bodies as the NRC and DOE requiring such an expensive, time-consuming, labor intensive effort, the immediate reaction is that LE must be a really powerful tool for decision making - maybe even the most powerful tool. There is even a strong temptation to equate this requirement with the statistician's request that every measurement be accompanied by a standard deviation so that the user knows how to evaluate the value of the measurement. There is a further temptation to assume that by calculating a LEID for every facility one creates a standard of comparison for all facilities. There is the ultimate temptation to believe that if the inventory difference or shipper/receiver difference is within the LE there is absolutely nothing to worry about. While some truth exists in each of the above premises, there is a real risk in blindly comparing LE's, quoting them to explain inventory differences, or even requiring them. This paper endeavors to explain what a l i m i t of error is, what it can be useful for, what it should be used for and who needs it for what purposes.