05 — Experimentation and Validation

Data preprocessing pipelines and validation gates that ensure data quality, distributional integrity, and experiment reproducibility before model training.

Notes

• Data Preprocessing — missing value imputation, scaling, categorical encoding, sklearn Pipelines, distribution transforms, outlier handling
• Data Validation — schema checks, KS test, PSI, Great Expectations, referential integrity, serving-time validation
• B Testing and Experiment Design — hypothesis testing, power analysis, sequential testing, common pitfalls
• Causal Inference — potential outcomes, DAGs, identification strategies, HTE estimation
• B Testing and Experiment Design — hypothesis testing, power analysis, sample size, sequential testing, Type I/II errors, SUTVA, common pitfalls

Links

← 04 — Feature Engineering → 06 — Interpretability and Communication

• Data Engineering (Production)
• Monitoring and Observability