Sensitivity analyses restricting to students assessed in all 15 terms reproduced the full-sample between-group comparisons¶

Claim · All claims
Evidence · 1 study · associational r2 · q2
Currency · newest study 2022 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Results for the subsample assessed in all 15 terms were qualitatively the same as the full sample for both achievement levels and growth patterns. → Johnson 2022

Evidence¶

Johnson 2022¶

Johnson, A. (2022). Achievement and growth for English Learners. Educational Policy, 37(4), 1044–1069. https://doi.org/10.1177/08959048211049419

q2 · i? · associational · r2

Sensitivity check addressing student attrition: the restricted subsample had slightly higher scores, but "the between-subgroup comparisons were qualitatively the same" for achievement ordering and year/summer growth patterns.

"Results for the subsample of students who tested in all 15 terms between the fall of kindergarten and the spring of 4th grade were similar to the full sample (see Appendix Tables A5 and A6)."

Discussion¶

Related Claims¶

• Systematic attrition differences mean COVID-19 impacts on student achievement are likely underestimated in fall 2020 MAP Growth results — the general claim this finding bears on, with the studies behind it
• The math discrepancy between NWEA and ERS recovery estimates is largely driven by differences in their samples of districts — related
• Supplemental analyses of I CAN Learn® in alternate second-year samples showed no effects that survived multiple-comparison adjustment — related
• Pooled effect sizes for deployment on mental health are robust across components of methodological quality in sensitivity analyses — a broader claim this one bears on
• The dashboard's cohort view restricts results to students who consistently tested across selected terms, and those outcomes may be positively biased — related
• The dashboard's cross-sectional view compares same-grade students across terms using the full sample of testers, but attrition is not random — related
• The intervention effects on mathematics are robust across model specifications and did not differ between RCT and COVID-affected implementation contexts — related
• Robustness checks with fixed-effect and first-difference models reproduce the three-way interaction; continuous-time growth curve modelling does not — related