Interpolation of experimental data without a theoretical model

A new approach is proposed for the interpolation of experimental data when a theoretical model is unavailable. The method is based on the minimization of a modified likelihood function incorporating a generalized smoothness test of the theoretical curve. The conditions for applicability of the method are a sufficiently accurate estimate of the statistical errors of the experimental data (assuming that the results fit a Gaussian distribution in accordance with the resulting rms deviations) and smoothness of the theoretical curve. A FORTRAN 77 program for the interpolation of experimental data has been written to implement the proposed algorithm. The computing time and roundoff error are determined as functions of the number of experimental points. (C) 1999 American Institute of Physics. [S1063- 7761(99)00209-7].