A Note on Histogram Approximation in Bayesian Density Estimation
A Andreev, Elja Arjas · 1996
Abstract In Bayesian density estimation, it is in practice necessary to restrict the space of density functions in some way, in order to arrive at an effectively finite parametrization. Here we consider piecewise constant functions as an approximating family. We show that if such functions are used to support an approximation of the “true” prior, then, under a set of natural conditions, the corresponding approximation will hold for the posterior.