POSTERIOR IDENTIFICATION OF HISTOGRAMS CONDITIONAL TO LOCAL DATA

Stochastic simulation techniques which do not depend on a back transfo rm step to reproduce a prior marginal cumulative distribution function (cdf) may lead to deviations from that distribution which are deemed unacceptable. This paper presents an algorithm to post process simulat ed realizations or any spatial distribution to reproduce the target cd f in the case of continuous variables or target proportions in the cas e of categorical variables, yet honoring the conditioning data. Valida tions conducted for both continuous and categorical cases show that, b y adjusting the value of a correction level parameter omega, the targe t cdf or proportions can be well reproduced without significant modifi cation of the spatial correlation patterns of the original simulated r ealizations.