Best Linear Unbiased Prediction

Abstract

Prediction problems arise naturally in life-testing experiments. For example, let us consider the life-test experiment described in Example 3 of Chapter 5. In this case, twenty three ball bearings were placed on a life-test and the data on the number of million revolutions before failure of each of these ball bearings were observed. The experiment itself was terminated as soon as the twentieth ball bearing failed with three ball bearings still surviving at the time of termination of the experiment. It is, therefore, natural for the experimenter to be interested in predicting the number of million revolutions before failure for the remaining three surviving bearings. In particular, the experimenter may be interested in predicting the very next failure or the very last failure.