← Every exhibit

Figure 2From “What is propensity score matching?”

A matched-estimate record

Write down the estimand, covariates, overlap, balance, unmeasured causes and benchmark before a matched estimate is used as an effect.

Cite Embed

Figure 2A matched-estimate record

A six-column worksheet for a matched estimate: the estimand and window, the covariates and whether they were measured before treatment, how many treated units were trimmed for lack of overlap, the standardized differences before and after matching, the unmeasured causes of treatment and outcome, and the experiment the estimate was held against or, failing one, the sensitivity analysis of how strong an unmeasured cause would have to be.EstimandEffect on whichtreated units,over whatwindow?CovariatesMeasured beforetreatment? Any itcould move?OverlapHow manytreated unitswere trimmed?BalanceStandardizeddifferences,before andafter?UnmeasuredWhat drivestreatment andoutcomeunrecorded?BenchmarkExperiment orsensitivityanalysis, andresult?Without an experiment, report an adjusted association and how strong an unmeasured cause would have to be toremove it.

Source: Author's decision worksheet. The estimand and covariate fields follow Rosenbaum and Rubin (1983), pp. 42 and 43, and Gordon et al. (2019), pp. 7 to 10; the overlap field follows Gordon et al. (2019), pp. 10, 12 and 16; the balance field follows Rosenbaum and Rubin (1983), pp. 48 and 52, and Gordon et al. (2019), pp. 16 to 19; the unmeasured and benchmark fields follow Gordon et al. (2019), pp. 22 to 27. Pages of Gordon et al. are those of the Articles in Advance version held. The fields are reader inputs; no company data is supplied.

PNG4:3, 1600 × 1200, with the title, the source line and the attribution set in the image isoglu-2026-what-is-propensity-score-matching-fig02.png

Source and evidence

Each line names a claim related to this exhibit, its source, and how it was checked.

  • B PSM-OWN-C3 A matched estimate is written down as one record: estimand and window, covariates measured before treatment, overlap and what trimming removed, balance on every covariate before and after, unmeasured causes, and the benchmark (an experiment or, failing one, a sensitivity analysis of how strong an unmeasured cause would have to be); without an experiment the record reports an adjusted association and that strength, never an effect; illustrated with a hypothetical B2B software firm's customer webinar series and expansion revenue, with an upgrade evaluation as the unmeasured cause and a holdout of invitations by lot for two quarters with a baseline quarter, provided uninvited accounts cannot attend, converted to the effect on attenders by the share attending as Gordon et al. did, which needs an argument that an invitation without attendance leaves expansion unchanged, since an invitation is itself a message to the account, or else the intention-to-treat effect is reported on its own (revised at G5 2026-10-15); untested.Author worksheet grounded in RR83-C4, RR83-C8, GZBC19-C1, GZBC19-C8, GZBC19-C9, GZBC19-C10 and GZBC19-C14

Grades: A, verified against the printed page of the primary source · B, primary source, text layer only · C, authoritative secondary · D, reported.

From the essay

Cite or embed

APA 7

Isoglu, S. (2026). A matched-estimate record [Figure]. isoglu.com. https://isoglu.com/insights/exhibits/isoglu-2026-what-is-propensity-score-matching-fig02/

BibTeX
@misc{isoglu-2026-what-is-propensity-score-matching-fig02,
  author = {Isoglu, Sinan},
  title = {{A matched-estimate record}},
  year = {2026},
  howpublished = {Figure 2, isoglu.com},
  url = {https://isoglu.com/insights/exhibits/isoglu-2026-what-is-propensity-score-matching-fig02/},
  note = {From the essay {What is propensity score matching?}}
}
Embed this exhibit

Copy this HTML to use the PNG exhibit on your site with its source line and a link back to the exhibit page.

<figure>
  <a href="https://isoglu.com/insights/exhibits/isoglu-2026-what-is-propensity-score-matching-fig02/">
    <img src="https://isoglu.com/social/isoglu-2026-what-is-propensity-score-matching-fig02.png" alt="Figure 2: A matched-estimate record" width="3200" height="2400">
  </a>
  <figcaption>Figure 2: A matched-estimate record. Source: Author&#39;s decision worksheet. The estimand and covariate fields follow Rosenbaum and Rubin (1983), pp. 42 and 43, and Gordon et al. (2019), pp. 7 to 10; the overlap field follows Gordon et al. (2019), pp. 10, 12 and 16; the balance field follows Rosenbaum and Rubin (1983), pp. 48 and 52, and Gordon et al. (2019), pp. 16 to 19; the unmeasured and benchmark fields follow Gordon et al. (2019), pp. 22 to 27. Pages of Gordon et al. are those of the Articles in Advance version held. The fields are reader inputs; no company data is supplied. <a href="https://isoglu.com/insights/exhibits/isoglu-2026-what-is-propensity-score-matching-fig02/">View exhibit</a>.</figcaption>
</figure>

Related exhibits

Figure 1Figure

Share of the raw comparison's bias left after stratifying on the propensity score, 14 Facebook checkout studies

What is propensity score matching?

Figure 2Figure

A nudge test record

What is a nudge?

Figure 2Figure

An analysis decision record

What is p-hacking?

Every exhibit

Contact

Bring the question you're working on.

Research, a guest lecture, or a commercial question.