A method for direct estimation of imprecision profiles, with reference to immunoassay data.

Abstract

A three-parameter model for directly estimating imprecision profiles from replicated immunoassay results is compared with a six-parameter indirect profile model obtained by the usual method of combining error in the raw response measurements with the slope of the standard curve. Direct estimation is likely to be less reliable when based on the limited data collected from a single assay, and may underestimate variability at high concentrations when many results are clustered at the upper end of the concentration range. However, at concentrations near the assay detection limit (often a region of particular interest), direct estimation is superior to the indirect method if a logistic or related function is used as the standard curve model. Direct estimation of imprecision profiles has useful application whenever the internal details of an assay system are not readily available, for example, in analysis of data collected in external surveys.