IEEE TRANSACTIONS ON RELIABILITY, VOL. R-20, NO. 4, NOVEMBER 1971 203

Say What You Mean

Two literary and logical difficulties often arise in discussing that the author defines the state of complete ignorance by thetechnical subjects: 1) not recognizing a definition as such, and prior distribution he chooses. The term has no otherthen treating it as a provable theorem; and 2) indirectly definition. (In practical applications to reliability, you shouldasserting that certainty will occur. b e concerned about describing your state of

The second is in principle easier to handle. When one knowledge-however meager it may be; it is a much moreasserts, "The resistor may or may not get hot," he really says profitable pursuit.)nothing since the statement is universally true. What he usually 3) Laws of Physics: Virtually all of the so-called laws ofmeans is either, "I am classifying resistors into two groups by physics are definitions of one of the parameters involved, notwhether or not they get hot," or "The resistor will not assertions about the universe. The "law" usually should benecessarily get hot." With a little thought, one can usually interpreted to mean that over a wide range of the variablesmake a statement that really says what is meant. involved, the parameter (defined by the equation) is

The first is less easy to handle. I will illustrate it by three reasonably constant. The trick in writing these equations-andexamples. in getting textbooks to name them after you-is to pick

1) The statement, "An unbiased coin has a probability of parameters to define that really do remain quite constant.1/2 for heads and 1/2 for tails," appears to be an observation Ohm's law is perhaps most well known to electrical engineers.about coins. But it is really only a repetition of the implicit E = IR defines R. Ohm's law says that for many substances thedefinition of unbiased coins. To avoid this difficulty, be sure R, so defined, is independent of E and I over wide ranges ofyou know what you mean when you use high-sounding their values. That this is often not true is attested to by thetechnical phrases. voltage coefficient which is often specified for precision

2) In Bayesian statistics, where probability is used as a resistors. Newton's law is also well known (in its nonrelativisticmeasure of degree of belief, many people want to represent form). F = ma defines m. Newton's second law then assertscomplete ignorance. Various prior distributions have been that the m, so defined, is independent of F and a, over wideused, depending on the properties of the parameter about ranges of both.which one wishes to represent the complete ignorance. Some Learn the word tautology: a statement that is true by virtueauthorities disagree on what those prior distributions should of its logical form alone-needless repetition of an idea. Manybe. But the term "complete ignorance" is not very well of your great ideas, and mine, under close examination turndefined, especially in an operational sense (it is like trying to out to be tautologies.find where -o is on a line). In practice, what really happens is