Standard assay procedures call for multiple replicates for the purpose of averaging random error from individual replicate scores to obtain a reliable estimate of the true score and for the purpose of specifying the degree of variability within the assay. In this paper we focus on the variability and magnitude of within-assay replicates in order to improve the estimate of the true score. A rule is presented for deciding (1) when individual replicates should be omitted from the computation of the assay's summary score or (2) when an entire sample of replicates should be discarded based on the pattern of replicate scores. It is suggested that this rule improves the reliability of the estimated assay score and this is supported by an illustrative analysis of lymphocyte proliferation assays.
- Assay replicate