SIMULTANEOUS MONTE-CARLO ZERO-VARIANCE ESTIMATES OF SEVERAL CORRELATED MEANS

Zero-variance biasing procedures are normally associated with estimati ng a single mean or ''tally.'' In particular a zero-variance solution occurs when every sampling is made proportional to the product of the true probability multiplied by the expected score (importance) subsequ ent to the sampling; i.e., the zero-variance sampling is importance we ighted. Because every tally has a different importance function, a zer o-variance biasing for one tally cannot be a zero-variance biasing far another tally (unless the tallies are perfectly correlated). The way to optimize the situation when the required tallies have positive corr elation is shown.