Index: IntHout 2014 - HKSJ vs DerSimonian-Laird Random-Effects Meta-Analysis

IntHout 2014 - HKSJ vs DerSimonian-Laird Random-Effects Meta-Analysis

Routing Summary

This folder fully captures IntHout, Ioannidis and Borm (2014), The Hartung-Knapp-Sidik-Jonkman method for random effects meta-analysis is straightforward and considerably outperforms the standard DerSimonian-Laird method (BMC Med. Res. Methodol. 14:25): formal DL and HKSJ definitions, the simulation design and Table 1 of error rates, the 689-analysis Cochrane comparison (Table 2), the DL-to-HKSJ conversion with both worked examples (Tables 3-6) and the discussion. 7 notes.

• Need the paper on one page, headline error rates, or the section map? → IntHout 2014 - HKSJ vs DerSimonian-Laird Random-Effects Meta-Analysis - Overview
• Need the DL and HKSJ formulas and the ${\text{var}}_{HKSJ}=q{\text{var}}_{DL}$ relation? → IntHout 2014 - HKSJ and DerSimonian-Laird Formal Specification
• Need the simulation design (scenarios, algorithms A-C)? → IntHout 2014 - Simulation Design
• Need Table 1 (min-max error rates) and the interpretation? → IntHout 2014 - Simulation Results and Error Rates
• Need the Cochrane 2012 empirical results (Table 2)? → IntHout 2014 - Cochrane Empirical Comparison
• Need to convert a DL forest plot to HKSJ, with the zinc and ALL examples and R code? → IntHout 2014 - Converting DL Results to HKSJ Worked Examples
• Need limitations, alternative methods and recommendations? → IntHout 2014 - Discussion, Limitations and Recommendations

Concept Map

Concept	Note	Type	Depends On	Key Result
DL variance and $z$-test	IntHout 2014 - HKSJ and DerSimonian-Laird Formal Specification	definition	Viechtbauer 2010 - tau2 Estimators, Q-Test and Heterogeneity Statistics	$1/\sum w^{\tau }$, normal
HKSJ variance and $t$-test	IntHout 2014 - HKSJ and DerSimonian-Laird Formal Specification	definition	DL	weighted residual variance, $t_{k-1}$
HKSJ as rescaled DL	IntHout 2014 - HKSJ and DerSimonian-Laird Formal Specification	theorem	both	${\text{var}}_{HKSJ}=q{\text{var}}_{DL}$
Size-mixture simulation	IntHout 2014 - Simulation Design	concept	I-squared	10,000 series per scenario
DL inflation	IntHout 2014 - Simulation Results and Error Rates	theorem	design	up to 30.9-33.7% at $k=2$
HKSJ near nominal	IntHout 2014 - Simulation Results and Error Rates	theorem	design	about 5%, at most about 10%
Table 1	IntHout 2014 - Simulation Results and Error Rates	definition	simulation	min-max by mixture, outcome, $k$
Cochrane comparison	IntHout 2014 - Cochrane Empirical Comparison	concept	HKSJ	25.1% of DL positives lost
Conversion recipe	IntHout 2014 - Converting DL Results to HKSJ Worked Examples	definition	HKSJ	zinc $p$: 0.04 to 0.16
Limitations	IntHout 2014 - Discussion, Limitations and Recommendations	theorem	all	caution for $\le 5$ unequal studies

Notes

• IntHout 2014 - HKSJ vs DerSimonian-Laird Random-Effects Meta-Analysis - Overview — CONTAINS: question, contributions, key results, full Source Map.
• IntHout 2014 - HKSJ and DerSimonian-Laird Formal Specification — CONTAINS: model, DL, HKSJ, scaling theorem and consequences, $I^2$, Python.
• IntHout 2014 - Simulation Design — CONTAINS: scenario grid, algorithms A-C, own replication and generator code.
• IntHout 2014 - Simulation Results and Error Rates — CONTAINS: Figure 1 reading, full Table 1, binary-outcome ranges.
• IntHout 2014 - Cochrane Empirical Comparison — CONTAINS: selection rules, Table 2, Thorlund corroboration, a flagged count inconsistency.
• IntHout 2014 - Converting DL Results to HKSJ Worked Examples — CONTAINS: recipe, Tables 3-6 step by step, R code, reproduction.
• IntHout 2014 - Discussion, Limitations and Recommendations — CONTAINS: evidence summary, caveats, alternative-method table, software status, conclusions.

Sources

• IntHout J, Ioannidis JPA, Borm GF (2014). BMC Medical Research Methodology 14:25. PDF: IntHout 2014 - Hartung-Knapp-Sidik-Jonkman Method for Random Effects Meta-Analysis

See Also

• Viechtbauer 2010 - Fitting Models with rma, Moderators and Knapp-Hartung, Roever 2020 - Overview and Motivation of Bayesian Meta-Analysis
• Multiple Testing Corrections, Power Analysis and Sample Size