A comparison of parameters below the limit of detection in geochemical analyses by substitution methods Primerjava ocenitev parametrov pod mejo določljivosti pri geokemičnih analizah z metodo nadomeščanja

Paper focuses on the analysis of geochemical data with values below the limit of detection (LOD). Such values are treated as text and are difficult to use in further calculations of mean, standard deviation and other statistical parameters. To estimate several methods for substitution of values below the LOD with fractions of LOD (zero, LOD/2, LOD/√2, LOD and no data values), a large dataset of generated values with normal and lognormal distributions was tested for different percent of censoring from 1 % to 50 %, plus the censored data of five selected geochemical parameters. Results indicate that the best substitution method is by LOD/√2, as it produces the smallest errors. The greatest errors are found for substitution methods with zero or no data. This is valid both for normally and lognormally distributed data. Median is not affected by most methods for censoring level below 50 %. For real geochemical parameters, the interpretation is more complex. For datasets with low amount of censoring (NO3, O2), the errors are small. For others (Sr, F, Mn) the errors are larger, as several LODs exist for each parameter and the LOD is sometimes larger than the mean value. Izvleček: V prispevku je predstavljena in analizirana problematika geokemičnih podatkov pod mejo določljivosti (MD). Ti so obravnavani kot tekst in se s težavo uporabljajo v nadaljnjih statističnih izračunih (povprečje, standardni odklon ipd). Primerjane so bile različne metode nadomeščanja vrednosti pod MD s petimi deleži: nič, MD/2,