Robust multiple method comparison and transformation
share
A generalization of Passing-Bablok regression for the simultaneous comparison of multiple methods is proposed. Possible applications include assay migration studies or interlaboratory trials. The method is shown to reduce to the usual Passing–Bablok estimator if only two methods are compared. It is close in spirit to reduced major axis regression, which is, however, not robust. To obtain a robust estimator, the major axis is replaced by the (hyper-)spherical median axis. The method is applied to the comparison of SARS-CoV-2 serological tests, bilirubin in neonates, and to a clinical test using different instruments, sample preparations and reagent lots. Plots, similar to the well known Bland-Altman plots, are developed for a posteriori checks of the assumed variance structure.
READ FULL TEXT