Does chemometrics work for matrix effects correction in X-ray fluorescence analysis?