Unbiased Estimators for Correlation Measurements

Abstract

Higher order correlation measurements involve multiple event averages which must run over unequal events to avoid statistical bias. We derive correction formulas for small event samples, where the bias is largest, and utilize the results to achieve savings in CPU time consumption for the star integral. Results from a simple model of correlations illustrate the utility and importance of these corrections. Single-event correlation measurements such as in galaxy distributions and envisaged at RHIC must take great care to avoid this unnecessary pitfall.

Comments: REVTeX, 22 pages, 3 hardcopy figures available on request, Phys Rev D (in press)