This paper provides extensions to the theory and the computational aspects of the Fellegi-Hold Model of Editing (JASA 1976). If implicit edits can be generated prior to editing, then error localization (finding the minimum number of fields to impute) can be quite rapid. In some situations, not all of the implicit edits can be generated because of the great number (> 10^30) of distinct edit patterns. The ideas in this paper are intended to determine more rapidly the approximate minimal number of fields to change in situations where not all implicit edits can be generated prior to editing. As a special case, the formal validity of Bankier’s Nearest-Neighbour Imputation Method (NIM) is demonstrated.
