- 01Data Profiling
- 02Strategy
- 03Execution
- 04Validation
Validation
The final module of the pipeline. aha runs five automatic quality checks on every harmonized variable and presents the results for you to review.
Overview
A summary panel at the top shows key statistics for the current session:
| Statistic | Description |
|---|---|
| Total Variables | The total number of variables in the unharmonized dataset. |
| Harmonized | The number of variables that passed all five validation checks. |
| Not Harmonized | The number that failed one or more checks and require attention. |
| Harmonization Rate | The percentage of variables that were successfully harmonized. |
| Not Mapped | The number of variables for which no standard mapping was found. |
The five validation checks
Each harmonized variable is evaluated against five criteria, and must pass all five to be marked as harmonized:
| # | Check | What aha evaluates |
|---|---|---|
| 1 | Concept Alignment | Do source and standard measure the same thing, at the same semantic level and units? Flags mismatches such as age vs. age group or categorical vs. continuous. |
| 2 | Data Type Compatibility | Are the data types compatible, and if not, does the plan define a valid conversion (e.g. numeric to categorical, datetime to year)? |
| 3 | Value / Category Coverage | Do mapped values cover all required standard categories, including edge cases such as multiracial or unknown? Unproducible categories are flagged. |
| 4 | Missing & Placeholders | Are all placeholder codes and missing-value strategies explicitly accounted for — sentinel values, nulls, and recoding? |
| 5 | Stats Sanity Check | Do output distributions and value counts look statistically reasonable — ranges and means for numeric, unique values and counts for categorical? |
The validation table
The main table lists every harmonized variable and its result. The Harmonized column shows TRUE for variables that passed all checks and FALSE for those that did not. The remaining columns give a brief explanation for each check, with a View more link for the full detail.
Use the sort button to bring failed variables (FALSE) to the top for easier prioritization.
Fixing failed variables
If one or more variables failed, use the Review not harmonized variables button at the bottom of the page. This returns you to Strategy & Planning with the list pre-filtered to only the variables that need attention. Once the plan is updated, re-run the pipeline form that point.
Pipeline processing time
The Pipeline Processing Time panel — opened by expanding the button of the same name — shows how many minutes were spent in each of the four modules during the session. It’s provided for transparency and helps estimate time for future projects.
Finalize the harmonization
Below the validation table, the Finalize & Save button completes the harmonization once you’re satisfied. The final harmonized CSV is saved to your workspace and you’re returned to the main menu.