Key points are not available for this paper at this time.
ABSTRACT This paper represents an attempt to formulate working definitions and principles by which the reliability of a given radioimmunoassay can be evaluated and compared with other radioimmunoassays. Statistical methods for the analysis of radioimmunoassay data are presented. The feasibility of utilizing the logit transformation has been tested empirically, and the advantages which accrue from analyzing data in this fashion are considered. The statistical transformations take into account the heteroscedasticity of the error inherent in a radioimmunoassay and the non-homogeneity extant in the confidence intervals about the estimates of the values of individual samples. By defining sensitivity as the smallest quantity of unlabeled hormone which can be measured with statistical certainty, and precision as the error in the determination of a hormone concentration at each particular point along the standard inhibition curve, it becomes possible to compare established assays in a meaningful fashion. Minimal criteria for establishing specificity, here regarded as freedom from interference by unwanted substances, are presented. Problems in the evaluation of relative and absolute accuracy are also discussed.
MIDGLEY et al. (Mon,) studied this question.