Nonparametric determination of energy spectra and mass composition of primary cosmic rays for slant depth

The data measured by the KASCADE (KArlsruhe Shower Core and Array DEtector) experiment are the basis for a multi-component analysis with the aim to determine the mass composition of the primary cosmic rays in the knee region. We discuss the methods used for estimating mass and energy of primary particles by utilizing neural network and nonparametric classification methods. By applying such techniques, measured data have been analyzed in an event-by-event mode and the mass and energy of individual EAS inducing particles are reconstructed. Results of all-particle energy spectra and relative abundances for different groups of primary particles are presented on basis of the electron and muon size data measured for different slant depths. The analyses of measured data indicate a transition to a heavier composition above a knee energy of ca. 5 PeV. It turns out that the mass composition depends on the particular set of observables (e.g. electron size N-e, truncated muon size N-mu(tr), hadron size N-h, most energetic hadron E-h(max),...) being considered simultaneously in the analysis. Though different sets of observables result in a qualitativly similar mass composition, quantitatively this leads to conspicuous differences. In this way the limitations of a particular interaction model are revealed and the necessity of detailed studies of correlations of EAS observables as a test of the hadronic interaction model is demonstrated.