Abstract: The data of 22 trace elements concentrations in human hair samples were obtained by ICP-AES and GFAAS. The variables which have significant influence on discriminating the sex are selected through the treatment of the concentration data by the variable dimension expansion and the variable selection methods. The discrimination plane figure with the good classification is obtained through the treatment of the data with selected variables by PLSmethod. The prediction models are built and used to distinguish the human sex according to the element concentrations data in human hair. The accuracy of the prediction is 81%.

Key words: Model selection, PLS regression, Trace element