MAXIMUM ENTROPY ESTIMATOR OF CONDITIONAL STOCHASTIC FIELDS

Abstract

This paper proposes a theory for evaluating the maximum entropy estimators and their associated entropy indices for conditional non-Gaussian translation stochastic fields when observation is made at some discrete points. Through analytical development and numerical examples of stochastic fields with the six types of distribution, kriging and maximum entropy techniques for spatial estimation are also compared. It was found that: 1) the maximum entropy estimate and conditional entropy are dependent on the values of observed data; and 2) the average conditional entropy has an independence on the observations, and is not larger than the average unconditional entropy.