Additive, multiplicative, and mixed analytical errors.

Abstract

Depending on the change in magnitude of analytical error occuring with change in concentration of analyte, two limiting types of error properties can be defined: additive and multiplicative. We investigated whether one of these two error types also characterizes the overall error of methods involving multiple procedural steps, or whether mixed error properties result in these cases. Using "open" quality-control data (i.e., and analyst identifies controls) from each of two hospitals and "blind" quality-control data from one hospital for 11 different assays, we found: (a) With current methodology, overall errors typically are mixed, though predominantly additive and multiplicative overall errors exist as well. (b) "Blinding" the quality-control system typically augments the multiplicative but not the additive error component.