inicioir al DIIbuscar
exploiting non-data information in statistical estimation
The major objective of this lecture is to lay the groundwork to include information in the formulation of statistical estimation problems beyond that coming from observations. A general framework is provided that covers a wide variety of situations when there is a poor, even a very poor, data base but some non-data information is available about the stochastic phenomenon. The approach is constructive in that it leads to efficient numerical procedures that can take advantage, sometimes with stunning results, of such additional information. The lecture will concentrate on the practical implications by means of a number of illustrations, sketch out the numerical procedure(s) and provide a brief outline of the underlying mathematical justification.