Application of partial least square (PLS) regression as calibration method was studied using as examples XRF analysis of galvanic bath and measurements of the thickness of nickel coating on steel base. Both linear and nonlinear PLS models were used. It was found that PLS regression performed not only the deconvolution of overlapped spectra, but could also correct interelement effects. This calibration procedure allows to convert nonselective measurements into selective information thus creating new possibilities of the application of radiometric methods based on low resolution X - or gamma-ray spectrometry.