4. The stochastic model

  • Redesign of the stochastic model now fully based on Least Squares and statistical theory.
  • Statistical hypothesis tests on normal distribution of the observations (Kolmogorov – Smirnow, χ^2 ).
  • Overall correlation analysis of the estimated parameters by the “Spectral Correlation Viewer (SCV)”.
  • The “Correlation RMSE Amplifier CRA” for verifying disadvantageous impacts of algebraic correlations of the LS parameters.
  • Generation of consistent residual spectra derived from the autocovariance function of the residuals to derive frequency dependent RMSE m0,i of arbitrary spectral domains di over the whole Nyquist interval.
  • Derivation of 95% confidence intervals for the frequency dependent RMSE m0,i and all estimated parameters.
  • Relative RMSEs (rRMSE) of the amplitude quotients and/or amplitudes for real and best circumstances as measures of achievable precisions for a specific station and observation record.
  • Quality criteria based on rRMSE, CRA, record lengths and signal strengths.