K-Means clustering of LIME feature-importance profiles identifies 2 responder and 4 non-responder subgroups within binary intervention-response classes¶
Claim · All claims
Evidence · 1 study · associationalr1·q2
Currency · newest study 2026 · checked against its sources 2026-10-26 (automated, 1 of 1 entry) · not reviewed by a person
Subclaims¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? Two responder subgroups (DBI = 0.81) and four non-responder subgroups (DBI = 0.48) were identified among 132 students with dyslexia using the Davies-Bouldin Index to select cluster solutions. → Tan 2026
Evidence¶
Tan 2026¶
Tan, A.S.-C., & Ali, F. (2026). Uncovering subgroup heterogeneity in dyslexia intervention outcomes using explainable machine learning. npj Science of Learning. https://doi.org/10.1038/s41539-026-00429-3
q2 · i? · associational · r1
K-Means clustering of LIME coefficients from a Gaussian Naive Bayes model (F1 = 0.72) applied to 132 dyslexia intervention participants. Lowest DBI solutions were chosen, meaning "the clusters were most distinct overall", supported by visual assessment. Subgroup sizes were 48/17 (responders) and 27/18/15/7 (non-responders).
"Two subgroups were identified under the responder group (DBI = 0.81), and four subgroups were identified under the non-responder group (DBI = 0.48)."
Discussion¶
Related Claims¶
- Longer gestation predicted literacy non-responsiveness while shorter gestation of 35-39 weeks predicted responsiveness — possibly the same claim (merge candidate)
- Strong visual memory for object recall predicted non-responsiveness, suggesting poor auditory working memory is a less significant predictor than strong visual memory — possibly the same claim (merge candidate)
- The multivariate omnibus test shows significant cognitive profile differences across the four reader groups — possibly the same claim (merge candidate)
- Exploratory K-means clustering identifies four heterogeneous student AI-user profiles differing in experience, competence, usefulness, trust, and control, but the solution's optimality and stability are not established — a broader claim this one bears on
- Cluster analysis identifies four tentative moral response profiles: Comfortable Adopters, Guilty Non-Users, Cautious Users, and Morally Distressed Avoiders — a broader claim this one bears on
- Latent profile analysis of 2015 PISA data identifies four collaborative problem-solver profiles among Chinese 15-year-old students — related
- Four distinct action-control belief profiles (agentic, moderate, avoidant, and mixed) emerge among Finnish ninth graders — a broader claim this one bears on
- Subgroup membership predicted distinct long-term literacy trajectories 68.2 months after intervention ended — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
lime-clustering-dyslexia-subgroups |
title |
K-Means clustering of LIME feature-importance profiles identifies 2 responder and 4 non-responder subgroups within binary intervention-response classes |
description |
K-Means clustering of LIME feature-importance profiles identifies 2 responder and 4 non-responder subgroups within binary intervention-response classes |
status |
review |
evidence_strength |
moderate |
newest_evidence |
2026 |
source_checked |
2026-10-26 |
generated |
by process:wiki-ingest on 2026-10-26 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — tan-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.