Utilising of mathematics-statistics methods concerning mechanic properties of heavy steel plates: dealing with the ill conditioned data

A problem of statistical modeling of some mechanical properties of heavy steel plates in the presence of colinearity of the observations is considered. The statistical analysis is based on the data gathered during the production process in the Steelworks "Czestochowa". The data appear to be statistically ill-conditioned and the models obtained with the help of the usual least squares method can be unsatisfactory. Thus to obtain regression models for the yield and tensile strength some Bayes-like estimators are proposed. The usefulness of the models obtained in the paper is examined with the help of the cross-validation techniques.