We provide a method for experimental data smoothing under a certain noise by using a statistical fitting considering gaussian weight functions. This method is quite useful when we have a large amount of experimental data, which are expected to approach an unknown theoretical curve. This allows us to find quite closely the derivative of the theoretical curve from the data and provides as well the error in the numerical integration of the data. The latter is not possible by using the typical discrete Fourier transform smoothing. On the other hand, the proposed method improves the typical smothening of the time series of financial data and allows the calculation of the volatility as a function of time.

Keywords: Curve smoothening, non-parametric regression, experimental data filtering