Method library WORKFLOW / 017

Research workflow/4 steps

Covariate balance for matching & weighting (cobalt)

@cobalt ✓ P D · Jun 29, 2026 · 801 views

@misc{cobalt,
  title        = {cobalt},
  author       = {Noah Greifer},
  howpublished = {\url{https://ngreifer.github.io/cobalt/}},
  note         = {Software / documentation}
}

In the margins · Peer notes@targeting_tara · Works with MatchIt, WeightIt, raw weights… one balance API for everything. Underrated. 2

The method workspace

Follow the analysis.

Select a node to explore the method

Method mapInput → 4 steps → output

START WITHInput data

ARRIVE ATResults

Read step 01

Inside the methodReasoning, equations & implementation

Data prep

Treatment W + covariates X

Step notesField note & discussion1

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

What happens here

Binary treatment plus the covariates to balance (Lalonde: age, educ, race, re74, re75).

Reads fromNot specified Feeds into#2

Key code

data("lalonde", package = "cobalt")

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 prep

Estimate weights / matches (WeightIt / MatchIt)

Step notesField note & discussion1

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

What happens here

Fit a propensity model; produce IPW weights or matched sets.

Reads from#1 Feeds into#3

Key code

library(WeightIt)
w <- weightit(treat ~ age + educ + race + re74 + re75, lalonde, estimand = "ATT")

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.

Diagnostic / pre-tests

[cobalt] Balance tables & Love plots — bal.tab()

Step notesField note & discussion1

A pre-flight check — run this before trusting any estimate downstream.

What happens here

bal.tab(): standardized mean differences, adjusted vs unadjusted.

Formula

\mathrm{SMD}=\dfrac{\bar X_{\text{treat}}-\bar X_{\text{ctrl}}}{\sqrt{(s^2_{\text{treat}}+s^2_{\text{ctrl}})/2}}

The estimator

Balance tables & Love plots — bal.tab() — Assess covariate balance before/after matching or weighting: standardized mean differences, KS stats, and the publication-ready Love plot.

OTHER Code @ HEAD ↗ Paper ↗

Reads from#2 Feeds into#4

Key code

bal.tab(w, un = TRUE)

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

love.plot()

Step notesField note & discussion1

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

What happens here

Love plot of |SMD| with the 0.1 balance threshold; show before vs after.

Reads from#3 Feeds intoNot specified

Key code

love.plot(w, thresholds = c(m = 0.1))

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.

Data contract

Inputs & data

A binary treatment and the covariates you'll adjust for (e.g. the Lalonde job-training data).

Show data format & exampleHide example

Format — one row per unit: treatment W ∈ {0,1} and covariates X.

  W   age  educ   race   re74    re75
  1    37    11   black     0       0
  0    30    12   white  4100    3800

Read the evidence

Results & interpretation

4 figures Workflow discussion 2

01FigureA Love plot of |standardized mean difference| per covariate, adjusted vs unadjusted, against the 0.1 threshold.

02FigureDistributional balance for a covariate before vs after adjustment.

03FigureCovariate balance across the sample, adjusted vs unadjusted.

04FigureOverlap / propensity distribution between treatment groups.

Figures reproduced from cobalt — Noah Greifer — unofficial community showcase; all credit to the original authors.

Context & interpretation

From the cobalt vignette (Lalonde). Estimate weights/matches, then check that adjustment actually balanced the covariates. Unofficial summary.

Peer notes

Discuss the workflow (2)

3

@calibrator_cleo · Jun 29, 2026

A causal estimate from observational data without a Love plot is a vibe, not evidence. cobalt makes the balance check trivial.

3

@aipw_amir · Jun 29, 2026

100%. I put love.plot() in every appendix now. Reviewers love it.
4

@targeting_tara · Jun 29, 2026

Works with MatchIt, WeightIt, raw weights… one balance API for everything. Underrated.