Downloaded from https://www.everyspec.com AMt" 70,1 where in the interval the true value Has, but or you arn quite uroe it lies in, or reasonably cloe, to that Interval. $(e) - eaq [- 2•(t)] (B-15a) The bluest reason that this concept is U) exp [+ 22(t)] (B-15b) satisfactory Is that often the region of un-J to)b certainty Is so large that the main message to the user Is that he knows much less than The estimates are mathematically well be. he hoped; or in a few canes, the region will hayed; e.g., It is easy to see that &, gL(t), be narrow enough that the user is quite iv(t) are Mways positive. satisifed, and would be even if the uncertainty/ were somewhat larger.
This derivation has been for the case where the sto~plngi rule is to stop aftei a fixed The upper and lower limits of the region length of time (cumulative test time of all of uncertainty for 2n Vdt) are units), Maximum Likelihood results are often the srane whether the stopping rule is for a In J,&(t) 0 In %(t) - 2;(t) (B-14a) "fixed time" or a "fixed number of failures". These estimates can be used regardless of In , at) in 1(t) + 2D(O) (B-14b) which kind of stopping rule is invoked. REFERENCE I. E. Parzen, Stochotic Poceune, Holden- Day, Inc. San Francisco, 1965. B-4- 2 .. 4 .-- -