NONHOMOGENEITY ANALYSIS USING BORROWED STRENGTH

This article develops a ''borrowed strength'' methodology for estimati ng local probability densities in a random field. Under a piecewise st ationarity condition similarities between the densities in different r egions of the field can be exploited by estimating a global mixture de nsity and imposing the parameter-space support of this borrowed streng th estimate on the local estimation problems. The local estimates are then used in an analysis of the homogeneity of the field, which is sho wn to benefit from the method of borrowing strength.