A robust microcomputer routine for the identification of outlying and influential points in radioimmunoassay standard curves.

This paper describes a suite of robust logit-log optimization routines for the fitting of a linear relation to laboratory data. The microcomputer routines assign a weighting value to each data point corresponding to the relative importance of that point. This procedure ensures that the routines identify and deal with any outlying data points. The routines are designed to generate the results of both a standard, unweighted least-squares algorithm and a robust, weighted algorithm thus providing the laboratory analyst with invaluable information concerning the data and the fitted relation. The routines are thoroughly tested on "perfect" and perturbed laboratory data.