Informative prior distributions for ELISA analyses

• DOI: 10.1093/biostatistics/kxu057
• Publication Date: 2015-01-11T01:10:56Z
• Authors: K. Klauenberg; M. Walzel; B. Ebert; C. Elster
• Keywords: Models, Statistical; Normal Distribution; Bayes Theorem; Enzyme-Linked Immunosorbent Assay; Biostatistics; 01 natural sciences; 3. Good health; Nonlinear Dynamics; Pregnancy; Calibration; Humans; Computer Simulation; Female; 0101 mathematics

Immunoassays are capable of measuring very small concentrations substances in solutions and have an immense range application. Enzyme-linked immunosorbent assay (ELISA) tests particular can detect the presence infection, drugs, or hormones (as home pregnancy test). Inference unknown concentration via ELISA usually involves a non-linear heteroscedastic regression subsequent prediction, which be carried out Bayesian framework. For such inference, we developing informative prior distributions based on extensive historical as well theoretical considerations. One consideration regards quality immunoassay leading to two practical requirements for applicability priors. Simulations show that additional information lead inferences robust reasonable perturbations model changes design data. On real data, is demonstrated across different laboratories, analytes laboratory equipment previous current ELISAs with sigmoid function. Consistency checks data (similar cross-validation) underpin adequacy suggested Altogether, new priors may improve estimation fulfill certain conditions, by extending analyses, decreasing uncertainty, giving more estimates. Future use these straightforward because explicit, closed-form expressions provided. This work encourages development application informative, yet general, other types immunoassays.