Nueva metodología para la detección de resultados inconsistentes en estudios interlaboratorios utilizando análisis de datos funcionales.

Abstract

This research proposes the development and implementation of a new methodology for outliers detection in interlaboratory studies, based on functional data analysis techniques and Mandel’s h and k statistics, which are in charge of the analysis of variability of study measurements. This work is developed in five chapters. The first one talks about the objectives and scope of the topic. Chapter 2 formally introduces interlaboratory studies and a general description of functional data analysis is provided, as well as certain definitions and techniques that can be applied to an interlaboratory study. Chapter 3 presents in detail the new methodology proposed for the detection of atypical laboratories in Interlaboratory Studies based on bootstrap resampling techniques, random projections and the functional extension of Mandel’s h and k statistics. From these statistics, the contrasting approach of the repeatability and reproducibility hypotheses for the functional case will be carried out. The new methodology has as a key idea transferring the resolution of a functional problem to a simpler case study, the real case. Chapter 4 presents the validation and application of the new proposed methodology. The validation consists in a simulation study of thermogravimetric curves, while the application is carried out in two real case studies. Additionally, a comparison and evaluation of the main results obtained against the results of previous studies related to the use of functional data in the framework of ILS is made. Finally, Chapter 5 shows the main conclusions and recommendations obtained during the development of the research.