STATISTICAL HANDBOOK FOR ASSESSING PIG CLASSIFICATION METHODS: Recommendations from the “EUPIGCLASS” project group

Acknowledgements: This work is part of the EU project Growth " EUPIGCLASS " (GRD1-1999-10914) partly funded through the 5 th Framework Programme , the Growth Generic Activity Measurement and Testing, and also partly through the Technological Development Programme from OFIVAL. Many thanks are due to Rainer Nagel and Juan Alvarez from the Pig and Poultry Office of the General Direction of Agriculture for their support and to Eli Olsen, from the Danish Meat Research Institute , for her excellent coordination of the EUPIGCLASS project. FOREWORD This publication is one of the outcomes of the Growth Project GRD1-1999-10914 " EUPIGCLASS ". This project aimed for standardisation of pig carcass classification in the EU through improved statistical procedures and new technological developments. It was coordinated by the Danish Meat Research Institute. The project was divided into three workpackages. The objective of the workpackage II was to solve the main statistical problems encountered in the application of the EC regulations for pig classification, to anticipate future problems and to form a basis to update the regulations. This statistical handbook was therefore written to provide useful and detailed documentation to people responsible for the national assessment of pig classification methods, both in the present and future EU countries. At the moment, this outcome of the EUPIGCLASS project has to be considered as a working document that could be used in discussions about changes in the EU regulations for pig classification. Depending these changes (probably decided upon in 2005), an update could be necessary before officially replacing the present version of the " Statistical Handbook for assessing pig classification methods " that was distributed at a EU Pigmeat Management Committee meeting in 2000. This new version of the handbook offers a considerable improvement over the 2000 version. The main proposal for an adaptation of the EC/2967/85 regulation is to replace the present error criterion " RMSE " by the " RMSEP " , that estimates the error of prediction and can be evaluated when Partial Least Squares (PLS) is used. The text is enhanced with many details and examples that will be of use to research workers and/or statisticians involved in pig classification.