Estimation of the response-error relationship in immunoassay

Area of Science:

• Biostatistics
• Analytical Chemistry
• Radioimmunoassay

Background:

• Accurate data reduction in immunoassays relies on understanding the response-error relationship.
• This relationship is crucial for developing appropriate weighting functions in statistical analyses.
• Commonly assumed functional forms for this relationship can impact analytical outcomes.

Purpose of the Study:

• To evaluate the statistical efficiency and reliability of different methods for estimating the response-error relationship in radioimmunoassay (RIA).
• To compare least-squares regression methods against maximum-likelihood based estimators.
• To assess the ability of estimators to guarantee positive predicted variances.

Main Methods:

• Generation of 50,000 simulated RIA response datasets using five common response-error functional forms.
• Estimation of parameters using three least-squares regression techniques.
• Application of three modified maximum-likelihood estimation methods.

Main Results:

• Maximum-likelihood estimators demonstrated superior statistical efficiency compared to least-squares methods.
• Two distinct maximum-likelihood estimators, despite differing computation times, yielded statistically indistinguishable results.
• Maximum-likelihood estimators reliably guaranteed positive predicted variances across the data range, unlike least-squares methods.

Conclusions:

• Maximum-likelihood estimation is the preferred statistical approach for analyzing RIA response-error relationships.
• The choice of estimator significantly impacts the validity and reliability of RIA data reduction.
• Ensuring positive predicted variances is critical for robust RIA data interpretation.