Record linkage results

1. Record linkage results Simulation example

2. Following probabilistic matching: For statistical analysis we require that correct variable value is carried into file of interest (FOI) Among candidate record matches we have (estimated) probabilities of a correct match and for each record the associated variable value This allows us to compute across all candidate records a distribution of the values of the variable This probability distn, for each record in the FOI (where match is equivocal) is then treated as a ‘prior’ and this is sampled from to ‘impute’ a value

3. Table 4: Standard probabilistic record linkage estimates for Time to infection with informative matching and different threshold choices. Mean standard errors in brackets. Number of simulated datasets = 100.

4. Table 5: Standard multiple imputation estimates for Time to infection with informative matching. Mean standard errors in brackets. Number of simulated datasets = 100.

5. Table 6: Prior informed multiple imputation estimates for Time to infection with informative matching. Mean standard errors in brackets. Number of simulated datasets = 100.