Method library WORKFLOW / 022

Research workflow/5 steps

Group & conditional effects with DoubleML (GATE / CATE)

@doubleml ✓ P D · Jun 29, 2026 · 906 views

Source DoubleML — Bach, Chernozhukov, Kurz & Spindler

@misc{doubleml,
  title        = {DoubleML},
  author       = {Bach and Chernozhukov and Kurz and Spindler},
  howpublished = {\url{https://docs.doubleml.org/}},
  note         = {Software / documentation}
}

In the margins · Peer notesA question, a caveat, a different perspective. Start a conversation here. 0

The method workspace

Follow the analysis.

Select a node to explore the method

Method mapInput → 5 steps → output

START WITHInput data

ARRIVE ATResults

Read step 01

Inside the methodReasoning, equations & implementation

Data prep

Build DoubleMLData (y, d, X)

Step notesField note & discussion1

Data preparation — shapes the raw inputs into what the estimator expects.

What happens here

Declare outcome, binary treatment and covariates.

Reads fromNot specified Feeds intoNot specified

Key code

dml_data = dml.DoubleMLData(df, y_col='y', d_cols='d', x_cols=X)

Discussion on this step (0)

StatsOtterDiscussion starter · Editorial note

Before using this step, make its assumptions and its output explicit. A useful peer note explains what you would check, what could go wrong, or what would change your interpretation.

Questions to explore

Which assumption matters most here, and what evidence could challenge it?
What would you check before passing this output to the next step?

Peer discussion

Bring your experience, a question, or a different perspective.

Estimation

Fit an Interactive Regression Model (IRM)

Step notesField note & discussion1

The core estimate — where the causal quantity itself is computed.

What happens here

Cross-fitted, Neyman-orthogonal IRM with ML nuisances — the per-unit signal for heterogeneity.

Formula

\psi(W;\theta,\eta)=g_1(X)-g_0(X)+\dfrac{D\,(Y-g_1(X))}{m(X)}-\dfrac{(1-D)(Y-g_0(X))}{1-m(X)}-\theta

Reads fromNot specified Feeds intoNot specified

Key code

dml_irm = dml.DoubleMLIRM(dml_data, ml_g, ml_m).fit()

Reference / docs ↗

Discussion on this step (0)

StatsOtterDiscussion starter · Editorial note

Before using this step, make its assumptions and its output explicit. A useful peer note explains what you would check, what could go wrong, or what would change your interpretation.

Questions to explore

Which assumption matters most here, and what evidence could challenge it?
What would you check before passing this output to the next step?

Peer discussion

Bring your experience, a question, or a different perspective.

Heterogeneity

Group Average Treatment Effects

Step notesField note & discussion1

Heterogeneity — who is affected, and by how much, not just on average.

What happens here

Average the orthogonal signal within each group to get a GATE per subgroup.

Formula

\theta_g=\mathbb{E}\!\left[\,\theta(X)\mid G=g\,\right]

Reads fromNot specified Feeds intoNot specified

Key code

gate = dml_irm.gate(groups=groups); gate.confint(joint=True)

Reference / docs ↗

Discussion on this step (0)

StatsOtterDiscussion starter · Editorial note

Before using this step, make its assumptions and its output explicit. A useful peer note explains what you would check, what could go wrong, or what would change your interpretation.

Questions to explore

Which assumption matters most here, and what evidence could challenge it?
What would you check before passing this output to the next step?

Peer discussion

Bring your experience, a question, or a different perspective.

Heterogeneity

CATE via a basis expansion

Step notesField note & discussion1

Heterogeneity — who is affected, and by how much, not just on average.

What happens here

Project the signal onto a spline basis to trace a smooth CATE in a covariate.

Formula

\theta(x)\approx \phi(x)^\top\beta,\qquad \theta(x)=\mathbb{E}\!\left[\,Y(1)-Y(0)\mid X=x\,\right]

Reads fromNot specified Feeds intoNot specified

Key code

cate = dml_irm.cate(basis); cate.confint(X_grid, joint=True)

Reference / docs ↗

Discussion on this step (0)

StatsOtterDiscussion starter · Editorial note

Before using this step, make its assumptions and its output explicit. A useful peer note explains what you would check, what could go wrong, or what would change your interpretation.

Questions to explore

Which assumption matters most here, and what evidence could challenge it?
What would you check before passing this output to the next step?

Peer discussion

Bring your experience, a question, or a different perspective.

Reporting

Plot with simultaneous bands

Step notesField note & discussion1

Reporting — turn the numbers into a figure or table a reader can act on.

What happens here

Show GATE bars and the CATE curve with joint (not pointwise) confidence bands.

Reads fromNot specified Feeds intoNot specified

Key code

# matplotlib: GATE bars + CATE curve with joint CIs

Discussion on this step (0)

StatsOtterDiscussion starter · Editorial note

Before using this step, make its assumptions and its output explicit. A useful peer note explains what you would check, what could go wrong, or what would change your interpretation.

Questions to explore

Which assumption matters most here, and what evidence could challenge it?
What would you check before passing this output to the next step?

Peer discussion

Bring your experience, a question, or a different perspective.

Data contract

Inputs & data

Outcome Y, binary treatment D, covariates X — plus the grouping variable(s) you want effects for.

Show data format & exampleHide example

Format — one row per unit: y, d ∈ {0,1}, covariates X.

  y     d   age  inc   educ
 1.2    1   41   38k   13
-0.3    0   53   21k   11

Read the evidence

Results & interpretation

3 figures Workflow discussion 0

01FigureGroup Average Treatment Effects — the effect within each subgroup, with joint confidence bands.

02FigureCATE estimated along a continuous covariate, with simultaneous bands.

03FigureA smooth CATE surface from a spline basis projection of the orthogonal signal.

Figures reproduced from DoubleML — Bach, Chernozhukov, Kurz & Spindler — unofficial community showcase; all credit to the original authors.

Context & interpretation

⚠️ Unofficial community showcase of a DoubleML example. Not affiliated with the authors; figures are from the public documentation. All credit to Bach, Chernozhukov, Kurz & Spindler.

Slice the average effect: Group Average Treatment Effects and a CATE surface from a debiased IRM, with simultaneous confidence bands.

Follow the analysis.

Inputs & data

Results & interpretation

Discuss the workflow (0)