Study of Intraclass Correlation Coefficient Method in a Measurement System

Abstract

There is the uncertainty of knowing with certainty the quality and appropriateness of the data obtained by a measurement; variation, bias, and calibration are some of the factors affecting the measurement systems, so the purpose of analyzing these systems is to ensure the validity of the results. This project involves measurement systems that have a need to get real results, i.e. that these systems try to determine the variation of the measurement, based on the equipment, piece, or the change of person and different measurements generated by the same person. The data are a representation of the process. The reliability of the collection of data must be ensured before doing any capability study, statistical process control (SPC). Without a validated measurement system, you can reach erroneous conclusions and act on the process the wrong way. It is therefore necessary to ensure that the variation is not of the measurement system used. Regarding the repeatability and reproducibility studies, one of the main actors is the automotive organization, AIAG (Automotive Industry Action Group), backed by their acceptance of the measurement system in a number of distinct categories. This is contradicted by Donald J. Wheeler who proposes a methodology using the intraclass correlation coefficient for correct measurement study. He references errors and inconsistencies in the AIAG method, with which measurement studies have been performed for many years. On this basis, the methods comparison study is performed. The Comparison of methods developed through simulations to study measurement from similar data, and using estimates, confidence intervals are created in order to compare the different results and determine if any of the methodologies are more likely to generate a correct measurement study.