Abstract:
An algorithm is considered which estimates the variance of a function which approximates an empirical dependence from a limited sample. The initial estimate is the difference between the mean value of the average risk functional and the estimate of the measurement error variance computed from the same sample.