Many sample value distributions are highly skewed and present variograms with high relative nugget effects and/or short-range structures. In consequence, significant deskewing of the histogram and variance reduction occurs when transforming between sample and block support, where blocks are of significantly larger volume than samples. When making estimates in the mining industry we usually wish to estimate on the basis of block support rather than sample support. The selective mining unit (SMU) refers to the minimum support upon which ore/waste allocation decisions can realistically and practically be made. The SMU is usually significantly smaller than the sampling grid dimensions, in particular at exploration/feasibility stages. In a non-linear estimate, for each large block (by convention called a panel') we estimate the proportion of SMU-sized parcels above a cut-off grade threshold. A series of proportions above cut-off defines the SMU distribution. Use of such non-linear estimates reduces distortion of grade-tonnage curves and provides a better basis for mine planning and economic decision-making. At present, Multiple Indicator Kriging (MIK) is the most widely used technique for performing estimation of recoverable resources. MIK is not the only approach available and it has some potentially serious limitations. The MIK technique is described and a number of theoretical and practical implementation issues are examined. In particular: MIK's assumptions about spatial distribution of grade; the behaviour of indicator variograms; application of change of support models; order relationship problems; and modelling the tail of the distribution. Some simple tests are described which allow us to determine the suitability of different estimators for a given deposit and criteria to identify a preferred estimator are also examined. Uniform Conditioning (UC), an alternative estimator which may be used instead of MIK, is described.