Multi-factor designs II. A design for identifying instruments with sample-to-sample carryover and drift

Abstract

We describe a nearly orthogonal two-level design that involves use of a weighted analysis to estimate drift and very low amounts of sample-to-sample carryover simultaneously. Identifying systematic errors from these sources is especially important for assays of analytes presenting a large range and with a medical decision point close to zero. The design is illustrated with data for thyrotropin, where, in one run with 32 samples, 0.08% carryover was detected in the presence of concentration-dependent negative drift.