data-validation

Run sanity checks on data aggregations by validating join keys and monitoring row counts before and after merging to detect join explosions. Apply checklist-driven cross-validation techniques to verify aggregation logic and ensure counts are not artificially inflated.

Denominator shifting occurs when the base population of a rate calculation changes unexpectedly across data segments. Catch it during data validation by applying denominator correctness checks to verify that base populations remain consistent and match the intended metric definitions.

Check for timezone mismatches by applying time-alignment validation to standardize timestamps across multiple data sources before analysis. Verify that all time-series data uses a consistent timezone to prevent skewed trends and incorrect chronological grouping.

Document data analyses for reproducibility by using structured templates to record data sources, metric definitions, methodology, assumptions, and versioning. This creates a clear audit trail that allows stakeholders to reproduce calculations and verify results independently.

Yes, you can detect selection bias by applying cross-validation techniques and reasonableness checks to identify skewed sample distributions. The checklist-driven validation process evaluates whether your data filtering logic inadvertently excludes relevant populations.

Cross-validation fails on null values and duplicate rows when pre-delivery QA skips completeness and deduplication checks. Resolve this by applying source verification, null handling, and deduplication steps before running aggregation logic and downstream calculations.