Box-Cox transformation on dataset from compositional studies of archaeological potteries

In archaeometric data set it was verified by statistical tests that some variables almost never follow a multivariate normal distribution using logarithmic transformation or other. This work presents a multivariate Box-Cox transformation for Mardia´s and Royton´s tests for a data set of fifty ceramic fragments from archaeological site Justino, Xingo, Brazil, and one clay sample collected near of the site. The samples were analyzed by instrumental neutron activation analysis, INAA. The study was made using companion to applied regression package from R software and was tested by Hotelling’s T 2 statistics.