Table 2. Summary of weights produced by record linkage using DOB as the blocking variable.
Linkage Run Name | String Matching | Weight Determination | Number of Weights | Minimum Weight | Maximum Weight | Pairs with Highest Weight | Pairs with Second Highest Weight | Pairs with Lowest Weight |
Pairs with Second Lowest Weight |
---|---|---|---|---|---|---|---|---|---|
R/EX/FS | Exact | Prob-FS | 8 | -12.3808 | 32.35027 | 30,536 | 24 | 176,066 | 189,273 |
R/EX/EM | Exact | Prob-EM | 8 | -17.7754 | 25.82314 | 30,536 | 24 | 176,066 | 189,273 |
R/EX/EPI | Exact | Prob-EPI | 8 | 0 | 1 | 30,536 | 24 | 176,066 | 189,273 |
MTB/EX/FS | Exact | Prob-FS | 9 | -12.3808 | 32.35027 | 30,536 | 24 | 176,056 | 10 |
MTB/EX/EM | Exact | Prob-EM | 9 | -17.7681 | 25.82333 | 30,536 | 24 | 176,056 | 10 |
MTB/EX/D | Exact | Det | 4 | 0 | 3 | 30,536 | 4,445 | 176,066 | 189,443 |
CU/EX/FS | Exact | Prob-FS | 9 | -12.3808 | 32.35027 | 30,536 | 24 | 176,048 | 10 |
LP/EX/FS | Exact | Prob-FS | 9 | -7.53877 | 12.78385 | 30,536 | 24 | 176,056 | 10 |
LP/EX/EM | Exact | Prob-EM | 9 | -7.53877 | 12.78385 | 30,536 | 24 | 176,056 | 10 |
R/INEX/FS | Inexact | Prob-FS | 8 | -12.3808 | 32.35027 | 31,619 | 25 | 176,018 | 189,095 |
R/INEX/EM | Inexact | Prob-EM | 121 | -18.5957 | 22.80771 | 30,536 | 3 | 176,018 | 189,095 |
MTB/INEX/FS | Inexact | Prob-FS | 64,273 | -12.3808 | 32.35027 | 30,536 | 3 | 5,691 | 1 |
MTB/INEX/EM | Inexact | Prob-EM | 64,273 | -17.7789 | 22.76115 | 30,536 | 3 | 5,691 | 1 |
MTB/INEX/D | Inexact | Det | 24,603 | 0 | 3 | 30,536 | 3 | 5,692 | 1 |
CU/INEX/FS | Inexact | Prob-FS | 1,492 | -12.3808 | 32.35027 | 30,554 | 3 | 173,105 | 2 |
LP/INEX/FS | Inexact | Prob-FS | 37 | 2.1 | 15.8 | 15 | 3 | a | a |
LP/INEX/EM | Inexact | Prob-EM | 37 | 2.1 | 15.8 | 15 | 3 | a | a |
DOB, date of birth; R, R package; MTB, Merge ToolBox; CU, Curtin University Probabilistic Linkage Engine; LP, Link Plus; Prob-FS, probabilistic, Fellegi-Sunter; Prob-EM, probabilistic, expectation-maximization; Prob-EPI, probabilistic, EpiLink; Det, deterministic.
a We were unable to recover negative weights for Link Plus with inexact string matching.