The Use of Control Variates in Monte Carlo Estimation of Power

Abstract

THE use of variance reduction techniques in simulation work has had little coverage in the statistical literature. Atkinson and Pearce (1976) suggest that one reason may be that ... except in simple or artificial examples it is difficult to obtain sufficient increase in precision to justify the additional complexities in programming. Applications of such methods which have led to worthwhile gains in efficiency are therefore of some interest. The variance reduction technique employed here is that of control variates. In this method an extra statistic, the control variate, is incorporated into the simulation study and some of its known sampling properties are used to increase the precision of the scheme. The rationale is as follows (Hammersley and