Constructing Ensembles of Pseudo-Experiments

Abstract

The frequentist interpretation of measurement results requires the specification of an ensemble of independent replications of the same experiment. For complex calculations of bias, coverage, significance, etc., this ensemble is often simulated by running Monte Carlo pseudo-experiments. In order to be valid, the latter must obey the Frequentist Principle and the Anticipation Criterion. We formulate these two principles and describe some of their consequences in relation to stopping rules, conditioning, and nuisance parameters. The discussion is illustrated with examples taken from high-energy physics.