Propensity score matching produced baseline-equivalent treatment and comparison groups on all measured variables¶

Claim · All claims
Evidence · 1 study · causal r2 · q3
Currency · newest study 2024 · checked against its sources 2026-10-22 (automated, 1 of 1 entry) · not reviewed by a person

Subclaims¶

q3 i? One-to-one optimal matching on demographics, fall 2020 MAP ELA scores, and 2021 behavior incidents yielded groups equivalent (g < 0.25) on all baseline variables for 653 matched pairs. → Gage 2024

Evidence¶

Gage 2024¶

Gage, N. A., Salomonson, K., Ballew, T., Clavenna-Deane, B., & Grasley-Boy, N. (2024). The Impact of Multi-Tiered Systems of Support (MTSS) on Student Attendance and Behavior. WestEd. https://eric.ed.gov/?id=ED655758

q3 · i? · causal · r2

Propensity scores were estimated by logistic regression on gender, race/ethnicity, low-income, disability, EL status, fall 2020 NWEA MAP ELA scores, and 2021 discipline incidents, followed by one-to-one optimal matching. Standardized mean differences for all baseline characteristics were below the WWC 0.25 threshold, with the largest at g = -0.24 (English Learner status).

"Table 1 provides the descriptive statistics for the 653 students in the three treatment schools and the 653 PSM students. We also calculated standard mean difference effect sizes (g) for each baseline measure to confirm equivalence. The students were equivalent (g < 0.25) on all variables."

Discussion¶

Related Claims¶

• In the full-sample analysis, baseline demographic and prior-achievement differences between treatment and matched comparison groups were minimal, with effect sizes below the WWC threshold — related
• Students in MTSS-coached schools had significantly fewer days absent in both implementation years — related
• Students in MTSS-coached schools were significantly less likely to have at least one behavior incident — related
• Using baseline data strongly predictive of outcome measures considerably reduces bias in nonexperimental estimates of charter school achievement effects — a broader claim this one bears on
• Propensity score stratification using multilevel models reduces selection bias from confounding variables and improves accuracy of charter school effect estimations — related
• A virtual comparison group matched on pretest, locale, FRL, and test dates achieved baseline equivalence with the treatment group (effect size = -0.03) — related
• Baseline equivalence held between conditions, but both groups scored far below the grade 8 proficiency cut score — related
• Baseline pre-assessment scores showed no statistical differences between treatment and control groups in either the matched or baseline sample. — related
• Propensity score matching produced comparison groups equivalent at baseline (within +/- 0.25 standard deviations) on all pretest achievement measures across the three QEDs — related
• Matched treatment and comparison grade-level clusters were balanced on observable baseline characteristics, with differences below a quarter of a standard deviation — related
• Learning-loss estimates are robust to placebo tests, propensity-score and maximum-entropy reweighting, and school and family fixed effects — related
• Conventional regression models cannot consistently estimate effects of time-varying treatments when time-varying confounders are affected by past treatment — related
• Across the full sample, students receiving the ASSISTments intervention performed comparably to their virtual comparison peers on end-of-year MAP Growth — related
• Implementing PBIS with at least Tier 1 fidelity showed no statistically significant effect on school disciplinary exclusions during the 2020–21 pandemic year — related
• Before intervention, treatment and control kindergartners did not differ on age, demographics, vocabulary, phoneme segmentation, letter knowledge, or reading — related