ABSTRACT
This chapter presents a summary of some of the power studies which have appeared in the literature and reference to some procedures which are powerful against specific alternatives. The history of general goodness-of-fit tests starts with the chi-squared test of Karl Pearson and includes such early omnibus procedures as the Kolmogorov-Smirnov and Cramer-von Mises tests. The original forms of these procedures are presently only of theoretical interest and are generally not used today with the exponential distribution since their power is lower than modern day tests and they are limited to use with simple hypotheses. Graphical procedures have the advantage that they allow the analyst to “see” the data and to make a subjective evaluation as to the appropriateness of the hypothesized model. Any peculiarities in the data can be readily detected and one is not faced simply with a decision to accept or reject a given model based on a single number.
