Computation of radioimmunoassay data using segmentally linearized standard curves.

Abstract A scheme for the calculation of the results of radioimmunoassays is described which does not require exact linearization of the standard curve. For the standards, a transformation related to the logit but without the extreme heteroscedacity of the logit is used. The resulting approximately linear standard curve is fitted, beginning at one end, as sequential straight-line segments. The fitted straight lines overlap and are each as long as is consistent with linearity. Methods are given for estimating the weighted hormone concentration and confidence limits for samples run at multiple dilutions. An approximate test for parallelism between dilutions of sample and standard is described. Differing nonspecific binding between samples may be allowed for. A computer program for performing the calculations is outlined.