POLYLINEAR REGRESSION-ANALYSIS IN RADIOCHEMISTRY

A number of radiochemical problems have been formulated in the framewo rk of polylinear regression analysis, which permits the use of convent ional mathematical methods for their solution. We have considered feat ures of the use of polylinear regression analysis for estimating the c ontributions of various sources to the atmospheric pollution, for stud ying irradiated nuclear fuel, for estimating concentrations from spect ral data, for measuring neutron fields of a nuclear reactor, for estim ating crystal lattice parameters from X-ray diffraction patterns, for interpreting data of X-ray fluorescence analysis, for estimating compl ex formation constants, and for analyzing results of radiometric measu rements. The problem of estimating the target parameters can be incorr ect at certain properties of the system under study. We showed the pos sibility of regularization by adding a fictitious set of data ''obtain ed'' from the orthogonal design. To estimate only a part of the parame ters under consideration, we used incomplete rank models. In this case , it is necessary to take into account the possibility of confounding estimates. An algorithm for evaluating the degree of confounding is pr esented which is realized using standard software of regression analys is.