Regression surfaces for estimated reflection coefficients

The distributions of the estimates of reflection coefficients depart considerably from normality, even using moderately long data records. As a result of this, the interdependencies among those estimates are nonlinear. The authors examine the nature of those nonlinear relationships by studying the regression surfaces of the estimated reflection coefficients, obtained using a classical nonGaussian model for the joint distribution of the latter parameters. Two theorems are then proved concerning the regression functions that allow one to analytically formulate the regression lines for the second-order case. Some conclusions are drawn about the asymptotic invariability of such lines and empirical validation is presented