Statistical design of data handling in environmental analysis

Abstract

In environmental analysis, owing to the huge number of analytical data, statistical evaluation is inevitable and must be carefully designed. Though the appropriate statistical test system is basically depending on the given analytical task, some general principles of the statistical design can be outlined, namely: i) before calculating mean value and standard deviation (SD), a preliminary test series must be performed for trend, distribution and outliers; ii) while the trueness of the measured mean values can be tested using appropriate certified reference materials (CRMs), even for a correct test, use of SD reference values are also necessary; iii) as the distribution may differ from the normal one, alternative non-parametric tests must be also built into the test system. The fulfillment of these requirements necessitates a flexible, adaptable statistical software, such as the DATAPRO 2.0 software recently developed.