Till Örebro universitet

The efficiency and quality of data editing processes are challenges for National Statistical Institutes (NSIs) in producing reliable official statistics. The traditional approach to data editing, heavily reliant on manual interventions, is resource-intensive and may introduce biases, impacting the overall accuracy of statistical estimates. This thesis aims to address these challenges by developing an in-novative editing framework based on probabilistic theory, allowing for a more resource-efficient editing process while providing accurate estimates of data quality. Furthermore, the thesis proposes an estimation procedure that accounts for various error sources, offering unbiased estimates of population parameters with appropri-ate measures of accuracy.

In addition to the introductory part, the thesis is structured around four key papers, each contributing to the overall objective of improving data editing and estimation processes in official statistics. Paper I presents a combined selective and probabilistic editing approach that maintains data quality while reducing resource demands. Paper II explores the integration of probabilistic editing with generalized regression (GREG) estimation, demonstrating improved accuracy in population parameter estimation. Paper III extends the framework to address nonresponse errors alongside measurement errors, using a three-phase sampling setup. Paper IV investigates the impact of various score functions in the probabilis-tic editing framework, emphasizing the importance of selecting effective score functions to minimize variance and improve estimate accuracy. Each paper contains, in addition to a theoretical part, an empirical section where concepts are numerically illustrated based on either real data or synthetic data.