Most current objective analysis schemes require a rather subjective pre-specification of “weight curves” in their interpolation formulae. This has the effect of preordaining, to a certain extent, the manner in which the variance will be distributed among the component perturbations. An objective analysis technique is presented wherein the data define their own “weight curves” and, consequently, their own variance partitioning procedures. The possible stochastic or periodic nature of the data field is discussed by means of the power spectrum. As a companion map to the final analysis, a map estimating the accuracy of the analysis at each grid point is presented. Abstract Most current objective analysis schemes require a rather subjective pre-specification of “weight curves” in their interpolation formulae. This has the effect of preordaining, to a certain extent, the manner in which the variance will be distributed among the component perturbations. An objective analysis technique is presented wherein the data define their own “weight curves” and, consequently, their own variance partitioning procedures. The possible stochastic or periodic nature of the data field is discussed by means of the power spectrum. As a companion map to the final analysis, a map estimating the accuracy of the analysis at each grid point is presented.