A new approach to estimate laboratory biases in interlaboratory comparisons is presented. Quantitative results as well as qualitative information from a survey (hybrid input data) are combined using a model that includes both a hierarchical regression model for biases and a structural equation model for the answers to the survey. The complete model applies when some latent variables explain simultaneously the measurement results and the answers to the survey.
Severine Demeyer
LNE, Laboratoire National de Métrologie et d'Essais