MétaCan
Menu
Back to cohort
Record W4392947957 · doi:10.1371/journal.pone.0298526

Do statistical heterogeneity methods impact the results of meta- analyses? A meta epidemiological study

2024· article· en· W4392947957 on OpenAlex

Why this work is in the frame

A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.

affAt least one author lists a Canadian institution in the pinned OpenAlex snapshot.

Bibliographic record

VenuePLoS ONE · 2024
Typearticle
Languageen
FieldDecision Sciences
TopicMeta-analysis and systematic reviews
Canadian institutionsUniversity of Alberta
Fundersnot available
KeywordsMeta-analysisRandom effects modelStatisticsConfidence intervalStudy heterogeneityPoolingMedicineRestricted maximum likelihoodInterquartile rangeMathematicsSample size determinationMaximum likelihoodInternal medicineComputer scienceArtificial intelligence

Abstract

fetched live from OpenAlex

BACKGROUND: Orthodontic systematic reviews (SRs) use different methods to pool the individual studies in a meta-analysis when indicated. However, the number of studies included in orthodontic meta-analyses is relatively small. This study aimed to evaluate the direction of estimate changes of orthodontic meta-analyses (MAs) using different between-study variance methods considering the level of heterogeneity when few trials were pooled. METHODS: Search and study selection: Systematic reviews (SRs) published over the last three years, from the 1st of January 2020 to the 31st of December 2022, in six main orthodontic journals with at least one MA pooling five or lesser primary studies were identified. Data collection and analysis: Data were extracted from each eligible MA, which was replicated in a random effect model using DerSimonian and Laird (DL), Paule-Mandel (PM), Restricted maximum-likelihood (REML), Hartung Knapp and Sidik Jonkman (HKSJ) methods. The results were reported using median and interquartile range (IQR) for continuous data and frequencies for categorical data and analyzed using non-parametric tests. The Boruta algorithm was used to assess the significant predictors for the significant change in the confidence interval between the different methods compared to the DL method, which was only feasible using the HKSJ method. RESULTS: 146 MAs were included, most applying the random effect model (n = 111; 76%) and pooling continuous data using mean difference (n = 121; 83%). The median number of studies was three (range 2, 4), and the overall statistical heterogeneity (I2 ranged from 0 to 99% with a median of 68%). Close to 60% of the significant findings became non-significant when HKSJ was applied compared to the DL method and when the heterogeneity was present I2>0%. On the other hand, 30.43% of the non-significant meta-analyses using the DL method became significant when HKSJ was used when the heterogeneity was absent I2 = 0%. CONCLUSION: Orthodontic MAs with few studies can produce different results based on the between-study variance method and the statistical heterogeneity level. Compared to DL, HKSJ method is overconservative when I2 is greater than 0% and may result in false positive findings when the heterogeneity is absent.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

Direct model labels (unvalidated)

Per-model category and study-design labels from the labeling rounds. They are machine output, unvalidated, and the disagreement between models ships as data. No study design here is MEDLINE-validated yet.

Model armCategoriesStudy designConfidence
gemmaMetaresearchMeta-epidemiology (broad)
Domain: Methods · Genre: Empirical
About the Canadian research system: no · About a Canadian topic: no
Meta-analysislow
gptMetaresearchMeta-epidemiology (narrow)Meta-epidemiology (broad)
Domain: Methods · Genre: Empirical
About the Canadian research system: no · About a Canadian topic: no
Meta-analysislow
models splitAgreement compares identical category sets and study designs across arms.

Full frame distilled prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.

metaresearch head score (Codex)0.391
metaresearch head score (Gemma)0.228
Version: codex-gemma-dda1882f352aValidation status: machine_predicted_unvalidated
Candidate categoriesMetaresearch, Insufficient payload (model declined to judge)
Consensus categoriesMetaresearch, Insufficient payload (model declined to judge)
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Meta-analysis · Consensus signal: Meta-analysis
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.178
Threshold uncertainty score1.000

Codex and Gemma teacher scores by category

CategoryCodexGemma
Metaresearch0.3910.228
Meta-epidemiology (narrow)0.0000.000
Meta-epidemiology (broad)0.0120.007
Bibliometrics0.0000.002
Science and technology studies0.0000.000
Scholarly communication0.0010.000
Open science0.0020.000
Research integrity0.0000.000
Insufficient payload (model declined to judge)0.0150.001

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.973
GPT teacher head0.700
Teacher spread0.273 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it