Resumen de Application of multivariate imputation of left-censored data in biomonitoring of radioisotopes
Z. Ziembik, A. Dołhańczuk Śródka
In biomonitoring living organisms, usually plants, are used in assessment of natural environment condition and in estimation of elements and chemical compounds concentration in an area. Determination of radioisotope’s activity concentration in a sample can provide valuable information. But one of the issues occurring during data interpretation is a result of low radioisotopes concentration in the material studied. Very often significant number of below MDA (Minimum Detectable Activity) communicates appear in output of signal analysis. As a result multivariate analysis of the data becomes impossible. To solve this problem data imputation method can be utilized. In our survey samples of moss, Ol soil horizon and shallow mineral layer were collected in Bory Stobrawskie forest, located in south-western part of Poland. In the samples activity concentrations of Cs-137, K-40, Pb-210, Pb-214, Bi-214, Ac-228, Th-231, and U-235 were determined using gamma spectrometry method. The data were grouped according to the material investigated. Activity concentrations of radioisotopes were recalculated to mass fractions. In general about 30The following functions were used for data imputation: multLN, lrDA, lrEM and multRepl. The functions were used with default parameters. Usually an influence of the imputation method on the data was limited. But for a variable with numerous MDAs influence of the method on generated concentrations was significant enough to disable appropriate data imputation.
