FPF2

FPF 2 is a second False Positive Filter that uses a four-dimensional weight table (mpi_wgt4dim) that can be configured. Generally, FPF2 is a better option than FPF.

The prerequisites are all or any subset of the following: birth year differences, birth date edit distance, gender, SSN edit distance, and name. The DOB, SSN, SEX, and XNM cmpspecodes must all be present, and they must all use the same comparison mode.

Reading FPF tables

In this table, dimension 1 is Name and Gender result.

Table 1. A sample FPF table
Dim 1 Index Name Result Gender Result
0 Missing Missing
1 Exact Missing
2 Partial Missing
3 Disagree Missing
4 Missing Agree
5 Exact Agree
6 Partial Agree
7 Disagree Agree
8 Missing Disagree
9 Exact Disagree
10 Partial Disagree
11 Disagree Disagree
  • Dim 2 is DOB Edit Distance:
    • Index 0 is DOB missing
    • Index 1 is edit distance 0
    • Index 2 is edit distance 1, and so on
  • Dim 3 is Birth Year Difference:
    • Index 0 – DOB missing
    • Index 1 - diff is 0-4
    • Index 2 - diff is 5-9
    • Index 3 - diff is 10-14
    • Index 4 - diff is 15-19
    • Index 5 - diff is 20 or greater
  • Dim 4 is SSN Edit Distance:
    • Index 0 – SSN missing
    • Index 1 – edit distance 0
    • Index 2 – edit distance 1, and so on
The mpi_wgt4dim table is used as the lookup for the FPF2. Here the index from the rules results is looked up and the corresponding weight is used. For example:
  • Entry 6, 2, 5, 4 is:
    • Partial Agreement on Name
    • Agreement on Gender
    • Edit Distance 1 on DOB
    • Difference 20 or greater on DOB
    • Edit Distance 3 on SSN