Application of the Hypothesis Analysis Method Using Cohen's Kappa Index to Measure the Agreement between Leather Sorters

摘要

One of the more sensitive activities carried out in a tannery, because of the impact the results could cause on the customer and on the tannery's business is, certainly, the sorting or grading of the hides in the different stages (wet blue or wet white, crust and finished). The surface defects affect the aesthetics appearance of leather and leather goods as well as the usable area. If the grade assigned to a hide by a sorter is lower than the one it really is, according to the agreement with the customer it will cause economical damage to the tannery. On the other hand, if the grade assigned is higher than the one it really is, the customer will be disappointed and claim. The detection of defects on the surface of natural materials such as leather is a difficult task because of the variety of shapes and textures, as well as in the quality and quantity of defective areas. The most common method applied in the leather industry to sort or grade hides is visual inspection. Notwithstanding the sorting or grading criteria could be well defined and agreed between the tannery and the customer, we can always wonder: if we change from one sorter to another, which will be the concordance level existing between them? In other words, how far the sorters will coincide on their evaluation? Will the result be a real or random one? Due to the variability between two or more sorters which is traditionally recognized as an important source of mistakes, from our point of view, it would be very useful to transform an attribute analysis into a variable one. The tanneries that supply leather to the automotive companies have to certify their Quality System according to QS 9000 or ISO/TS 16949. One of the requirements in that norm and technical specification is to apply the Automotive Industry Action Group (AIAG) reference Manual "Measurement Systems Analysis" which describes the study of the measurement systems by attributes. In this paper, we describe the practical way on which the agreement between sorters is measured and the application of the hypothesis analysis method, cross tabulation, calculating Kappa coefficient initially proposed by Cohen, which is a statistical concordance rating measure for qualitative (categorical items) between two sorters.