A two-stage estimation procedure when there may be spurious observations

Abstract

In laboratories where blood samples are analyzed, it is common to use a two-stage procedure for estimating a reaction time. The procedure consists of making two independent determinations and if they differ by more than d units, a third determination is made. In the case of a third determination, one of the first two is discarded and the estimate is the average of the retained observations. Assuming that each observation is, with some probability, a spurious one the bias and the mean square error of the estimator are derived. The problem of choosing the d value is considered. The effects of changes in the d value on the bias and the mean square error are examined.