Record Linkage Techniques -- 1997 Proceedings of an International Workshop and Exposition (1999) / Chapter Skim
Currently Skimming:
Chapter 12 Tutorial on Record Linkage
Chapter 12 Tutorial on Record Linkage
Pages 455-480
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 455... ...
1 ·1 Record Linkag Techniques 1 997 Tutorial on Record Linkage Authors: Martha E Fair anciPatricia Whitricige, Statistics Can acia 455
|
From page 457... ...
· Gett ng the data ready for linicage p~p~ssing Basic operations in a typical record linkage project · Searching - looking for the correct linkage · Decision making · Grouping · Post-processing of files aRer lint - ge Kline (3) · Tncks of Be trade · Examples of applications in health, business, and agriculture · References where to get more inforrnadon · Glossary of terms · Question period - interest of audience Introduction · Deficit on of record linkage · Stadshcal uses of record linkage · Administrative uses of record linkage Deterministic hnicage · Probabilistic linkage Record Linkage Future i~i =]
|
From page 458... ...
~ Matching and eliminade~n of duplicate Small area studies Collaborative records Events People ~ Methods for hard-to~numerato Cross-seabonal Longitudinal | popolabons Record Linkage-Todays Sltuadon (1) · Shift from paper~ased systems to electronic · Optical imaging of source documents · Generalized systems · Suite of compare products · Commercial softwares
|
From page 459... ...
. A theory of record linkage.
|
From page 460... ...
· The efficiency of the record linkage operation depends on how well the items selected for comparison satisfy this standard.
|
From page 461... ...
__ Vocabulary -- Basic Terms (3) · Possibly linked pairs 0 Gray area · Unilnkedinonlink pairs o Unmatched · Global weights · Frequency weights · Discnminaffng power · Specific discriminating power Nune: Den a' R ~IDES Zecha~ Orvil Outstay 1897 03 23 MarUnique Zed~arb~s Ondl Burley 1887 05 25 Mardn~que Dow linbtS~5 If there are ~_...
|
From page 462... ...
Set C has N x Nb record pairs (3x2=6)
|
From page 463... ...
Record Linkage Techniques 1997 Slides Presentation (cont'd) Rules and Thresholds (3)
|
From page 464... ...
Fair and Whitridge Slides Presentation (cont'd) Number of Links; vs.
|
From page 465... ...
Record Linkage Techniques-1997 Slides Presentation (cont'd)
|
From page 466... ...
Birth Date -- 11rS _ , ..
|
From page 467... ...
SEARCHING Phase · Objective is to search for pairs that are to Iy linked · Possibly apply early rejection rules e.g not one item other than the pocket identifiers agree · Dec de on the most efficient order of comparisons e.g quick cutoff · Speaty nobles and weights to be used in the comparisons TOPIC THREE Dec~sion-makin~ Phase Weights · Creating comparison rules · Setting thresholds ~ Manual resolution - optional Usirlg The Discrmiinatin'2 Power of Items (1) -c, · Agree,disagree,missing · Agree, disagree, partial agreements · Agree, disagree, partial agreements with global weights · Agree, disagree, partial agreements using frequency weights Conditional agreements .
|
From page 468... ...
Comp~g Surname_ TIPS · Phonetic coding · Partial agreements · String comparators · Maiden versus married names · Watch out for titles - Sr.
|
From page 469... ...
. Mapping Documentation of Me Process · One to one ~ Data dichonanes · One to many ~ Record layouts · Many to one ~ Flow diagrams · Many to many · Histograms of weights · Threshold settings · Conflict resolution ~ Rules and weights used · Manual resolution ~ Analysis ale · Updates Muld-pass Linkages · Decide on Me number of passes required · Each pass should have different blocking criteria · Choose blocking items that do not overlap in order to pick up the missing links not achieved on an earlier pass · Examples: NYSIlS code and sex code Bird, date and first forename Errors, Their Sources and Magnitudes · Blocking information.
|
From page 470... ...
) O EMUS O WIT ~ ;r' |
From page 471... ...
Slides Presentation (cont'd) Use of Record Linkage ~ Cancer Re~tries · Treason or cancer repasses · Maintaining cancer registries · Dead clearance of cancer registries · Evaluadug He quality of registries · Ascerta~nrnent of new death certificate only cases · Replacing or partially replacing active follow-up of patient · Grrying out cohort studies · Follow-up of clinical trials and scr~g prod Advantages of Record Linkage in Cancer Registries · Reduces respondent burden · improves accusal · Reduced follow-up costs · Refines detection and measurement of mortality and cancer rates for particular cohorts USE OF RECORD LINKAGE IN BUILDING, MAINTAINING ANI)
|
From page 472... ...
Provide information teddy Bet Safely Stan~dar~s ~ Ontario miners study 3 Assist With Health Promos Advises ~ National Breast Screening program Somo~cono~c Gradient In Mortality If_ k-~9~ · Manitoba Cenbe for Head Policy and Evaluation · Statistics Canada · Mortality and health care util~affon described in relation to sac oeconomic status · Measure mortality and use of health care seMces RECORD LINKAGE AGRICULTlJPE AND BUSINESS APPLICATIONS Long-Term Medical FoBow-up Results (21 Only INK 4. Assist Task Forays.
|
From page 473... ...
~TRODUCIlON · Record linkage techniques developed pnmanly for matching indwicluals · Some aspens similar fw businesses, some very different · Special challenges are present in rural areas and for agricultural population · Incentives to match admin data rather than run new surveys MATCHING WEARABLES · Commonly available: name ('ndividual or business) address phone number ~ industrial classification (type of business)
|
From page 474... ...
· Unincorporated businesses may be owned by many partners · Individuals may be involved in more than one business ARE: In~m~ ElDIIOHmeDt · Unix machine 0 GRLS (Generalized Record Linkage System, developed at Statistics Canada) - Beta release · Approximately 280,000 incoming Census records to be matched against 400,000 Farm Register records CENSUS OF AGRICULTURE Linkage Process · Need to link Census farms to Farm Register · 3 step process: exact match, probabilistic match, Men manual resolution · Exact match - incoming Census fauns matched in SOL - indudes prey and post-processors CHALLENGES: CENSUS OF AGRICULIIJRE Structural Problems (cons)
|
From page 475... ...
Record Linkage Techniques-1997 Slides Presentation (cont'd) CENSUS OF AGRICULTURE Challenges · Census records are farms, most matching fields at farm operator level · Many farmers involved in more than one farm; many farms operated by more than one farmer; most fangs unincorporated · No data available to help match or resolve multiple matches CENSUS OF AGRICULTURE Chin (I)
|
From page 476... ...
Co" Building, Stat OR Tuso~'Puturc Om - ,Oe~do K1~0T6 Summary · Record linkage soRware development · Quality of data files · Uniform ciassificabon standards ~ Analysis of data · Analysis of data - incorporating uncertainty due to linkage shoos: (6t3)
|
From page 477... ...
GLOBAL FREQUENCY RATIOS for agreement outcomes and partial agreement outcomes are often subsequently converted to this value-specific counterpart during the linkage process. The conversion is accomplished by means of an adjustment upwards where the agreement portion of the identifier has a rare value, and an adjustment downwards where the value is common.
|
From page 478... ...
As relating to a particular outcome from He comparison of a given identifier it is synonymous why He FREQUENCY RATIO for Hat outcome. As relating to the accumulated FREQUENCY RATIOS for a given record pair it refers to He overall RELATIVE ODDS.
|
From page 479... ...
The use of the logarithm is merely a convenience when doing the arithmetic; it does no affect the logic except to make it appear more complicated. The term 'WEIGHT" has therefore been employed sparingly in this book.
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.