Skip to main content

Table 1 Traditional match weights, match weights incorporating dependence between identifiers, and attribute-specific match weights according to agreement pattern {date of birth, sex, postcode}. Record pairs with no agreement on any identifiers, or where only sex agreed (agreement patterns {000} and {010}), were assumed to be non-matches and excluded

From: Utilising identifier error variation in linkage of large administrative data sources

  

Agreement pattern {date of birth, sex, postcode}

001

100

011

101

110

111

N Matches

 

1

21

18

12

15,924

14,009

N Non-matches

 

259

414,307

248

4

415,888

10

Match probabilitya

0.0039

0.0001

0.0726

0.7500

0.0369

0.9993

Traditional match weight

5.3

−1.0

9.6

14.9

8.6

23.7

Match weight assuming dependence

−0.5

−1.2

9.3

17.8

8.6

27.6

Attribute-specific match weight:

      

Sex

Female

−1.7

−1.7

8.7

17.3

8.7

27.7

Male

0.4

−0.9

9.7

18.1

8.5

27.5

Age

0–1 years

−0.5

0.1

8.6

18.5

9.2

27.6

5–6 years

−0.2

−2.0

9.6

18.2

7.9

28.0

18–19 years

−1.5

−2.6

9.5

16.2

8.3

27.2

Ethnicity

Missing

2.7

0.3

10.8

19.5

8.5

27.6

White

−1.3

−1.6

8.8

17.4

8.6

27.6

Mixed

1.8

−0.4

10.9

19.4

8.8

28.6

Asian

−0.4

−2.3

10.4

16.9

8.6

27.8

Black

1.6

0.9

9.6

19.1

8.9

27.1

Other

1.4

−0.2

10.5

19.0

8.8

28.1

Organisational-specific match weight (mean)

5.7

1.3

12.3

20.7

8.1

25.4

  1. aMatch probability = N matches/Total record pairs