Answers are grounded in this paper and cite their sources.
Introduction. Part 1 The basics and their application: including probabilistic linkage, betting odds and frequency ratios, using global and specific discriminating powers. Part 2 Exploiting more of the discriminating power: including what to do with missing identifiers, comparing names, years of birth, places of birth, geographical identifiers and marital status, linked and unlinkable pairs, some pitfalls. Part 3 Saving central processor time: including blocking the files, the preliminary rejections, the comparison sequence and its early cut-off, the application of value-specific discriminating powers. Part 4 Organizing the product: including calculating absolute versus relative odds, setting an optimum threshold, grouping the matched pairs. Part 5 Recapitulation and further thoughts. Appendices: A - definitions. B - derivations. C - typical identifier frequencies. D - linking special disease registers. E - errors, their sources and magnitudes. F - calculating the outcome frequencies for a non-existent file of unlinkable pairs. G - constructing a file of randomly matched unlinkable pairs. H - details of phonetic coding systems for names. I - design of a computer linkage system. Bibliography. Index.
No takes yet. Share an insight, caveat, or question.
Winkler et al. (1989) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: