# Final Assessment: Causal Inference

## Assignment

Write an identification memo for an observational causal question.

Your memo must state the estimand, draw the causal model in words, choose an identification strategy, name assumptions, plan diagnostics, and explain interpretation limits.

## Required Artifact

Submit a one- to two-page identification memo.

Minimum sections:

- decision context
- estimand
- DAG in words
- confounders, mediators, colliders, and selection mechanisms
- identification strategy
- assumptions
- diagnostics
- sensitivity analysis
- interpretation limits
- strongest critique

## Rubric

| Criterion | Strong | Needs revision |
|---|---|---|
| Estimand | Names treatment, control, population, outcome, horizon, and ATE/ATT/LATE/CATE | Describes a relationship without a target effect |
| Causal model | Clearly distinguishes confounders from mediators and colliders | Treats all variables as controls |
| Strategy | Matches method to design assumptions | Chooses method by popularity |
| Assumptions | Makes exchangeability, positivity, SUTVA, and measurement assumptions explicit | Leaves assumptions implicit |
| Diagnostics | Plans balance, overlap, placebo, pre-trend, or manipulation checks as appropriate | Uses diagnostics as proof rather than stress tests |
| Interpretation | States who the estimate applies to and where it may fail | Generalizes beyond the data |
| Critique | Names the strongest threat to identification | Only lists minor limitations |

## Pass Criteria

Pass if the memo makes the causal claim criticizable and clearly separates evidence, assumptions, and decision implications.

## Submission Checklist

- [ ] The estimand is complete.
- [ ] The DAG is described before the estimator.
- [ ] Post-treatment variables are flagged.
- [ ] Identification assumptions are named.
- [ ] Diagnostics are appropriate to the strategy.
- [ ] Sensitivity analysis is planned.
- [ ] Interpretation limits are explicit.

## Certificate Language

Completed Causal Inference by producing an identification memo with estimand, causal model, assumptions, diagnostics, sensitivity plan, and interpretation limits.