A/B Testing & Causal Inference
Follow the data, decisions, feedback, and validation boundaries before writing the full system.
How to read this diagram
Read left to right for the forward path: raw information becomes a representation, passes through the project’s main computational ideas, and produces an output that can be measured. Then follow the feedback path back toward the trainable or decision-making components.
Power analysis
Causal estimation separates treatment effect from ordinary variation by stating an identification assumption. Power, parallel trends, interference, pre-treatment fit, and multiple testing must be checked before interpreting a p-value or effect estimate.
Boundary check: document its accepted input, output shape, mutable state, failure modes, and the metric that proves this stage is correct before connecting it downstream.
DiD
Causal estimation separates treatment effect from ordinary variation by stating an identification assumption. Power, parallel trends, interference, pre-treatment fit, and multiple testing must be checked before interpreting a p-value or effect estimate.
Boundary check: document its accepted input, output shape, mutable state, failure modes, and the metric that proves this stage is correct before connecting it downstream.
Synthetic control
Causal estimation separates treatment effect from ordinary variation by stating an identification assumption. Power, parallel trends, interference, pre-treatment fit, and multiple testing must be checked before interpreting a p-value or effect estimate.
Boundary check: document its accepted input, output shape, mutable state, failure modes, and the metric that proves this stage is correct before connecting it downstream.
Architecture review checklist
- Every arrow has a documented shape, dtype, unit, or schema.
- Training and evaluation paths cannot leak information into each other.
- Randomness is seeded and captured in experiment metadata.
- Expensive stages expose timing, memory, throughput, and error metrics.
- Each feedback loop has a stop condition and a rollback strategy.
- Small reference implementations exist for numerical comparisons.