SOME 2ND THOUGHTS ON ERROR STATEMENTS

The problems connected with the evaluation of the uncertainty to be associated with the result of an experimental measurement have recently been the object of much renewed interest. In this review some of the conflicting suggestions are analyzed critically and an attempt is made to arrive at objective criteria for judging their respective merits and drawbacks. In particular, the need traditionally claimed for à clear-cut distinction between the random and the systematic contributions is questioned, as is the usefulness of expressing systematic uncertainties in terms of a maximum limit. It is argued that a possible unification can be based upon an approach in which all uncertainties are given in the form of estimated standard deviations, allowing then a general application of the law of error propagation. In this way, a number of awkward problems arising in more conventional prescriptions for handling uncertainties can be avoided. © 1979.