Prediction and imputation in ISEE: tools for more efficient use of combined data sources

Abstract

The development of an Integrated System for Editing and Estimation (ISEE) is an important part in Statistics Norway's strategic plans for improvement of statistical production processes and more efficient use of available data sources. ISEE is organized in applications for different processing functions, and implemented in a service-oriented IT structure. Two of the applications are DYNAREV (for editing and imputation) and STRUKTUR (for estimation of population aggregates). With the two processes of editing and estimation being fully integrated in ISEE, a producer of statistics is now in a much better position to implement the so-called top-down approach to editing, because the effect on the estimates of the population totals due to any changes made to the data can be examined instantly. In this paper we provide an overview of the various tools for prediction and imputation in ISEE. Some of these are well in place whereas others are still being developed. Our main focus is on the construction of a statistical register. We propose and discuss a triple-goal criterion, and assess the alternative imputation methods in accordance, and finally outline a method that is potentially capable of satisfying these needs.