Mathematical Geology, Vol. 26, No. 3, 1994 Comparative Performance of Indicator Algorithms for Modeling Conditional Probability Distribution Functions ~ P. Goovaerts 2 This paper compares the performance of four algorithms (full indicator cokriging, adjacent cutoffs indicator cokriging, multiple indicator kriging, median indicator kriging) for modeling conditional cumulative distribution functions (ccdf). The latter three algorithms tire approximations to the theoretically better full indicator cokriging in the sense that they disregard cross-covariances be- tween some indicator variables or they consider that all covariances are proportional to the same function. Comparative perfi~rmance is assessed using a reference soil data set that includes 2649 locations at which both topsoil copper and cobalt were measured. For all practical purposes, indicator cokriging does not perform better than the other simpler algorithms which invoh'e less variogram modeling effort and smaller computational cost. Furthermore, the number of order relation deviations is found to be higher fiJr cokriging algorithms, especially when constraints on the kriging weights are applied. KEY WORDS: indicator kriging, conditional probability, order relation deviation, E-type esti- mate. soil geochemistry. INTRODUCTION Most recent developments in geostatistical theory have focused on the modeling of uncertainty prevailing at any unsampled location. The priority is no longer at getting a "best" estimate for the unsampled value but at modeling the cu- mulative distribution function (cdf) expressing the uncertainty about that value given the information available. Such conditional cdfs (ccdfs) can be established using a variety of algorithms that are classified as parametric and nonparametric. The difference between the two classes resides in the way the ccdfs are modeled. In the parametric approach, an a priori analytical model is assumed (e.g., mul- tivariate normal or lognormal). The conditional distributions are then fully char- acterized by a few parameters (mean, variance) that can be estimated easily by L Received 15 June 1993; accepted 1 November 1993. 2Stanlbrd University, Geological and Environmental Sciences Department, Stanford, California 94305. 389 0882 81211941041.~)-03895n7 ool I :c 1994 Inlemalmnal Ass(~:iallon for IvlalhcmatLcal Gr