Estimation of the heteroscedastic noise in large data arrays Original Research Article

An approach has been developed to estimate the uncertainties in experimental data that follows a heteroscedastic model. The method presented is based on a hypothesis that the size of data is sufficiently large such that the data values over a limited domain have approximately homoscedastic variance. The implementation of the procedure has two steps. The first step is to estimate an approximate error for each data element. The second step is to estimate the standard deviation for the data points using the errors estimated for the neighboring data. The related mathematical theory is presented, and several simulated data examples are used to illustrate this approach.