The Match Designer provides information in various formats
to help you evaluate test pass results. These formats include the
test results data grid, charts, statistics, column-level metadata,
and the Weight Comparison window. This information
helps you assess the current test results and plan for subsequent
match passes.
About this task
For example, after you test the match passes, the following
process shows the steps that you might take to evaluate the test results:
- Click the Test Results pane and the Pass Statistics tab
to view the test results.
- Click the Total Statistics tab to view
the combined statistics for all the match passes.
- In the Test Results pane, select a subset of records in the data
grid and right-click Compare Weights. For those
selected records, view the weight that the matching columns contribute
to the composite match weight.
These areas of the Match Designer enable you to perform
the following actions:
- Explore the match results
- View statistics about the test run
- Create charts
- Establish baseline results and compare previous test results to
current test results
- View the column details such as agreement and disagreement weights
by right-clicking a column
- View the weight that the matching columns contribute to the composite
match weight
Procedure
Procedure
- To help determine cutoff values, sort the Test Results
data grid in descending order by weight.
- Examine the sort results.
- If you begin to question the match results at a certain
weight range, set the clerical cutoff to that weight.
- If it seems that questionable matches are occurring at
higher weight ranges, give the match columns higher m probabilities.
Large m probabilities like 0.99999 are appropriate for very important
columns. This value assigns a large penalty when values in this column
disagree.
- Right-click the Test Results data grid and include residual
records in the output display. If it seems that some of these records
should have been classified as matches, consider lowering your match
cutoff, clerical cutoff, or both.
- Optional: After you evaluate the results, you
might want to modify some elements of one or more passes, test the
passes, and then evaluate them again.
Results
When you are satisfied with the results of all passes in a
match specification, you are ready to use it in an Unduplicate Match
or Reference Match stage.
IBM InfoSphere DataStage and InfoSphere QualityStage, Version 8.5
This topic is also in the IBM InfoSphere QualityStage User's Guide.
