We formulate selective editing as a combinatorial optimization problem whose solution establishes which sampled units contain influential errors and thus must undergo interactive editing within a generic editing and imputation strategy. This optimization problem arises naturally from considerations on editing resources savings and estimates accuracy control. Cross-sectional auxiliary information is taken into account through linear mixed models assisting the construction of the problem's feasibility region. We provide a general algorithm for the univariate version of this problem, i.e. for editing one single variable. By applying this proposal to each questionnaire variable we illustrate its use upon the Spanish industrial turnover index and industrial new orders received index surveys. A reduction of interactive editing with a controllably increase of estimates error is observed.
Palabras clave / Key words: Selective editing, combinatorial optimization, auxiliary information, linear mixed modelsExploiting auxiliary information: selective editing as a combinatorial optimization problem (Pdf 374 KB)