Trial Sequential Analysis
Adjust cumulative meta-analysis for repeated significance testing — is the evidence conclusive, or is more information needed (Wetterslev et al. 2008)?
- Family
- Meta-analysis
- Engine
- Lacuna native (Python)
When to use it
Use it when you have an effects table and need a defensible pooled estimate.
Takes meta-analysis out of a calculator output and binds it to a decision trail: effect size, heterogeneity, model selection and provenance.
Limitations and scope
- No standalone figure ships for this method yet; its results are reported as tables and interpreted findings.
- Results are conditional on the studies, corpus and parameters you supply, and on the chosen model. The analysis summarises the evidence as reported; it does not establish causation.
- Every figure carries its source, and every run is reproducible bit-for-bit from the same inputs, parameters and engine version.
In this family
- Univariate Meta-AnalysisPool one effect size per study into a single summary estimate with heterogeneity diagnostics.
- Multilevel Meta-AnalysisTwo- or three-level models for dependent effect sizes nested within studies or labs.
- GLMM Meta-AnalysisGeneralised linear mixed models for binary/count outcomes without normal approximation.
- Network Meta-AnalysisMixed treatment comparison across ≥3 interventions with a connected evidence network.
- Diagnostic Test AccuracyBivariate sensitivity/specificity model with a summary ROC curve.
- Dose–Response Meta-AnalysisModel the shape of an exposure–outcome relationship across dose levels.
- P-Uniform*Bias-corrected effect estimate robust to selective publication.
- P-Curve AnalysisTest the significant p-values for evidential value vs p-hacking via the p-curve's right-skew (Simonsohn et al. 2014).
- MA Power AnalysisProspective / retrospective power for a meta-analytic design specification.
Run this analysis
The definition above is open. The live engine, its parameters and the provenance-sealed report run inside your workspace.