FlexRL - Flexible Record Linkage and Linked Data Quality Assessment

Probabilistically link records that refer to the same entities across two data sources without a unique identifier, using partially identifying variables such as product code, brand, category, birth year, sex or postal code. 'FlexRL' implements a Stochastic Expectation Maximisation (StEM) approach to Record Linkage (Robach et al., 2025, <doi:10.1093/jrsssc/qlaf016>). The model accounts for registration errors (missing values and mistakes) and for variables that change over time, enforces one-to-one assignment, and has a low memory footprint. The package also provides tools for inference on linked data: two estimators of the false discovery proportion of a linkage (Robach et al., 2025, <doi:10.1002/sim.70292>), based on linkage scores and on synthetic data, and diagnostics comparing the linked sample with the source data. These tools also apply on the linkage output of other record linkage packages.