brycewang-stanford/causal-inference-mixtape

Causal Inference: The Mixtape — Code Skill

Practitioner-oriented causal inference skill built from Scott Cunningham's Causal Inference: The Mixtape repository. Covers 10 identification strategies with ready-to-run code templates in Python, R, and Stata.

---

Methods Covered

| Method | Python | R | Stata | Reference | |--------|--------|---|-------|-----------| | OLS / Regression | statsmodels | estimatr | reg/reghdfe | references/method-patterns.md §1 | | Difference-in-Differences | statsmodels + C() | lfe/fixest | xtreg/reghdfe | references/method-patterns.md §2 | | Event Study (Dynamic DiD) | manual lead/lag | estimatr | reghdfe | references/method-patterns.md §3 | | Staggered DiD / TWFE | statsmodels | bacondecomp | bacondecomp | references/method-patterns.md §4 | | Regression Discontinuity | statsmodels polynomial | rdrobust | rdplot/rdrobust | references/method-patterns.md §5 | | Instrumental Variables | linearmodels IV2SLS | AER/ivreg | ivregress 2sls | references/method-patterns.md §6 | | Synthetic Control | rpy2 → R Synth | Synth + SCtools | synth | references/method-patterns.md §7 | | Matching / PSM / IPW | manual logit + weights | MatchIt + Zelig | teffects/cem | references/method-patterns.md §8 | | DAGs / Collider Bias | dagitty (conceptual) | dagitty/ggdag | — | references/method-patterns.md §9 | | Randomization Inference | permutation loop | ri2 | ritest | references/method-patterns.md §10 |

---

Core Workflow

Implement a Causal Method

1. Identify the method from the table above
2. Load the appropriate template from references/method-patterns.md
3. Adapt variable names, fixed effects, and clustering to the user's data
4. Add robustness checks (parallel trends for DiD, McCrary for RDD, first-stage F for IV)

Choose the Right Language

| Scenario | Recommendation | |----------|---------------| | ML pipeline integration | Python (statsmodels + linearmodels) | | Synthetic Control | R (Synth package) or Stata (synth) — Python lacks mature implementation | | Bacon decomposition | R (bacondecomp) or Stata — no Python equivalent | | Publication-ready tables | Stata (outreg2/esttab) or R (stargazer/modelsummary) | | Coarsened Exact Matching | Stata (cem) or R (MatchIt) — no Python equivalent | | Quick prototyping | Python with statsmodels |

Cross-Language Equivalents

| Task | Python | R | Stata | |------|--------|---|-------| | OLS with robust SE | smf.ols().fit(cov_type='HC1') | lm_robust() | reg y x, robust | | Cluster SE | fit(cov_type='cluster', cov_kwds={'groups': g}) | felm(y ~ x | 0 | 0 | cluster) | reg y x, cluster(id) | | Two-way FE | C(id) + C(time) in formula | felm(y ~ x | id + time) | reghdfe y x, absorb(id time) | | IV / 2SLS | IV2SLS.from_formula('y ~ 1 + exog + [endog ~ inst]') | ivreg(y ~ exog | inst) | ivregress 2sls y exog (endog = inst) | | DiD | C(treat)*C(post) | treat:post in formula | did_multiplegt or interaction |

---

Key Python Patterns

DiD with Cluster-Robust SE

import statsmodels.formula.api as smf

model = smf.ols('y ~ C(treated)*C(post) + controls', data=df)
results = model.fit(cov_type='cluster', cov_kwds={'groups': df['firm_id']})

Event Study (Lead/Lag)

# Create relative time dummies
for k in range(-4, 5):
    col = f'rel_{k}' if k >= 0 else f'rel_m{abs(k)}'
    df[col] = (df['relative_time'] == k).astype(int)

# Drop t=-1 as reference
formula = 'y ~ ' + ' + '.join([c for c in rel_cols if c != 'rel_m1']) + ' + C(id) + C(year)'

IV / 2SLS

from linearmodels.iv import IV2SLS

model = IV2SLS.from_formula('y ~ 1 + exog + [endog ~ instrument]', data=df)
results = model.fit(cov_type='clustered', clusters=df['cluster_var'])

---

Robustness Check Patterns

| Method | Required Checks | |--------|----------------| | DiD | Parallel trends (event study plot), placebo treatment dates | | RDD | McCrary density test, bandwidth robustness (half/double IK optimal), polynomial robustness | | IV | First-stage F > 10, exclusion restriction argument, over-identification test | | Synthetic Control | Pre-treatment RMSPE, placebo distribution, leave-one-out | | Matching | Covariate balance table, caliper sensitivity |

---

Common Pitfalls

1. TWFE with staggered treatment — standard two-way FE is biased when treatment timing varies. Use Bacon decomposition or Sun & Abraham / Callaway & Sant'Anna estimators.
2. Synthetic Control with many treated units — the Synth package handles one treated unit. For multiple, use augmented synthetic control or stacked approach.
3. RDD without McCrary test — always test for manipulation at the cutoff before estimating.
4. IV weak instruments — report first-stage F-statistic. Below 10 indicates weak instrument bias.
5. Python Synth gap — no mature Python Synth package exists. Use rpy2 to call R's Synth from Python.

---

Additional Resources

Reference Files

• references/method-patterns.md — Detailed code templates for all 10 methods with full examples
• references/r-stata-comparison.md — Cross-language package comparison and method coverage gaps

Prompt Files

• prompts/01-implement-method.md — Copy-paste prompt for implementing any causal method
• prompts/02-robustness-checks.md — Copy-paste prompt for generating robustness check code