Merge Passes -- Guidelines

1. Maximum acceptable false positive rate. The table of best matches probably will not be representative of all matches and might have very few entries.

2. Number of imputations for missing links. To test sensitivity of linkage results, create 10 imputations and compare 1--5 to 6--10.

3. Coefficient of Variation for Total Matches. If you have two estimates (Min, Max) then set Total Matches = (Min + Max) / 2 and Coefficient of Variation = ((Max - Min) / 6) / Total Matches.

4. Size of random sample of record pairs. Smaller samples run faster and larger samples give better estimates.

5. Number of MCMC iterations for burn in. To test sensitivity, specify 5 and check parameter estimates in the Markov Chain table.

6. Seed for random number generator. To test sensitivity, run with Seed = 1, save the results, and rerun with Seed = 2, etc.

7. Method for optimizing imputed pairs in sets. If you want to apply a different algorithm, start with the Merged Pairs tables and/or the Imputed Pairs tables.