On the estimation of an experimental function