Un algoritmo de suavización para curvas empiricas

Abstract

There are several methods for smoothing empirical functions. A standard method consists in the minimization of the quadratic norms for both, the vector of residuals, and the vector of the p-th derivative of the function to be estimated. Here, the contribution of the nuIl-space of the derivative functional is considered and Green's functions are used for inverting this operator. The algorithm contains global and local means of controlling the degree of fitting. The procedure is applied to the standard seismological problem of smoothing travel-time tabIcs and the results are compared with two other methods mentioned in the litera tu re, namely, a eubic spline scheme with variable knots and the summary method of Jeffreys.