Abstract
We have developed practical methods for evaluating the magnitude of the random errors in radioimmunoassay dose--response variables, and the relationship between this error and position on the dose--response curve. This is important: to obtain appropriate weights for each point on the dose--response curve when utilizing least-squares curve-fitting methods; to evaluate whether the standards and the unknowns are subject to error of the same magnitude; for quality-control purposes; and to study the sources of errors in radioimmunoassay. Both standards and unknowns in radioimmunoassays for cAMP and cGMP were analyzed in triplicate. The same mean (Y), sample standard deviation, sy, and variance (2-y) of the response variable were calculated for each dose level. The relationship between s 2-y and y was calculated utilizing several models. Results for standards and unknowns from several assays were pooled, and a curve smoothing procedure was used to minimize random sampling errors. This pooling increased the reliability of the analysis, and confirmed the presence of the theoretically predicted nonuniformity of variance. Thus, the calculation of results from these radioimmunoassays should utilize a weighted least-squares curve-fitting program. These analyses have been computerized, and can be used as a "pre-processor" for programs for routine analysis of results of radioimmunoassay.
Extrinsic Least Squares Regression with Closed-Form Solution on Product Grassmann Manifold for Video-Based Recognition
