The invention relates to a method of analyzing spectral data for the selection of a calibration model, relating spectra of a substance to a physical or chemical parameter of the substance, over a predetermined range of the physical or chemical parameter, comprising the steps: a) capturing spectral data of the substance with respective values of the physical or chemical parameter over the predetermined range, b) creating a plurality of calibration models using the captured spectral data in dependence upon the values of the physical or chemical parameter based on the calibration data using statistical resampling methods, c) calculating tolerance intervals of the results at each reference level for each calibration model, and d) displaying the tolerance intervals at each reference level over the predetermined range for each calibration model. In this way, a possibility for analyzing spectra data is provided which is useful in spectroscopic applications for automated calibration model selection and makes analytical interpretation and evaluation easier and more accurate.