Applied Causal Inference & Decision Science

Turning messy transactions into tested panels with treatment, timing, outcomes and pre-period covariates.

• Days 1–3Causal vocabulary, Polars cleaning, DuckDB and the customer × week panel.
• Days 4–7Ingesting API data with dlt and modelling the panel in dbt.
• Days 8–10Data contracts for leakage and treatment timing, orchestration with Prefect.

Stating the question precisely and making every causal assumption explicit.

• Days 11–13Potential outcomes, estimands (ATE, ATT, CATE, LATE) and randomization as the benchmark.
• Days 14–17DAGs, backdoor and frontdoor criteria, colliders and bad controls.
• Days 18–20Encoding assumptions in DoWhy and defending a DAG in writing.

Designing, running and analysing randomized experiments well.

• Days 21–23Test design, power and sample size, stratified randomization and balance.
• Days 24–27Regression analysis, CUPED, multiple testing, peeking and sequential tests.
• Days 28–30Interference, switchback and geo experiments, and the free-shipping capstone.

Estimating effects without randomization when confounders are measured, or an instrument exists.

• Days 31–33Regression adjustment, propensity scores and matching.
• Days 34–36Inverse probability weighting, overlap and doubly robust (AIPW) estimators.
• Days 37–40Instrumental variables, 2SLS, weak instruments and LATE.

Evaluating rollouts and campaigns when you can't randomize.

• Days 41–44Difference-in-differences, event studies and modern staggered estimators.
• Days 45–47Synthetic control, synthetic DiD and Bayesian structural time series.
• Days 48–50Regression discontinuity and placebo tests.

Using machine learning to estimate effects in high dimensions and find who responds.

• Days 51–54Orthogonalization, cross-fitting and DoubleML for partially linear, interactive and IV models.
• Days 55–57Meta-learners, causal forests and validating heterogeneous effects.
• Days 58–60Price elasticities with DML and instruments, by segment and category.

Stress-testing estimates and turning them into targeting policies and choices under uncertainty.

• Days 61–63Sensitivity to hidden confounding, robustness values, E-values, refutation tests and specification curves.
• Days 64–66Uplift modelling, policy learning and cost-benefit analysis.
• Days 67–70Decisions under uncertainty, value of information and the decision memo.

Putting causal estimates in front of decision-makers as tools they can use.

• Days 71–73A versioned parameters file, a scenario API with FastAPI, scenario design and uncertainty for executives.
• Days 74–76The Streamlit simulator, counterfactual charts, tests, CI, a container and a smoke test.
• Days 77–80Monitoring for effect drift, when to re-run the experiment, and the final capstone defence.