[Metameters for the statistical treatment of data expressed in percentage values].

Medical or biochemical data are often expressed as percentages or proportions. Since the sampling error of percentage values does not usually follow a normal (Laplace-Gauss) model, statistical significance tests may give unsatisfactory results, when applied to the original data; however, the fitting to a normal distribution can be improved by choosing suitable metameters. The Author proposes the transformation of percentages to a hyperbolic function zp, based on Fisher's z, which is almost normally distributed. The performances of this transformation are checked, with special reference to medical laboratory problems.