Statistical Modeling of Within-Laboratory Precision Using a Hierarchical Bayesian Approach.

The article focuses on developing a hierarchical Bayesian modeling (HBM) approach to estimate within-laboratory precision—defined as repeatability and intermediate precision—in food analysis using duplicate measurement data from routine laboratory tests. By analyzing over 300 instances across various analytes measured by methods such as spectroscopy, gas chromatography (GC), and high-performance liquid chromatography (HPLC), the study establishes a nonlinear regression model incorporating a constant term, a concentration-dependent term, and lognormal-distributed random effects to account for sample heterogeneity. The model fits well except for moisture, a method-defined analyte, and generally predicts standard deviations within the Horwitz ratio criteria, with some high-sensitivity detector cases showing better precision. The authors propose using this model for internal quality control and measurement uncertainty estimation without requiring detailed consideration of sample matrices, recommending at least 300 duplicate analyses over a wide concentration range for robust parameter estimation.