Revised Passing-Bablok Regression Method for Model Comparison

Abstract

Type-II regression models are used to compare more than one method that makes the same measurement. The Passing-Bablok regression method, which is one of them, is non-parametric and can give more successful results than other comparison methods, especially when there are outliers. In this study, innovations in the calculations of slope and intercept parameters used in the traditional Passing-Bablok method are proposed. Instead of the median parameter used in the classical model, the use of the trimean parameter was suggested and the model parameter estimates were adjusted accordingly. The proposed new model and classical model predictions were compared on 15 different data sets, 8 of which were simulations. It has been determined that the proposed new model calculations contain fewer errors than the results of the classical method.