Speaker
Description
metacart 3.0 integrates regression and classification trees into the framework of meta-analysis to perform exploratory moderator analysis. Meta-regression trees identify, based on the study characteristics and their interactions, subgroups of studies that maximize within-subgroup homogeneity of effect sizes. To avoid overfitting, the resulting tree is pruned using cross-validation. Finally, subgroup meta-analysis is applied to test the significance of moderator effects.
Version 3.0 introduces several methodological advances targeting two core challenges: tree construction stability and potential bias in the statistical inference in the identified subgroups. For stability, a smooth sigmoid surrogate strategy can be used to replace a greedy search and can be complemented with a newly incorporated look-ahead strategy to improve the splitting procedure. For improved inference, a permutation test is added to evaluate the significance of $Q-$between and a bootstrap procedure is included to correct for bias in the effect size estimates of the subgroups.
Simulation study results confirm that these new features yield satisfactory false discovery rates (< .05), improve the residual heterogeneity estimation, and reduce bias in the estimates of effect sizes in the identified subgroups.
Finally, new diagnostic plots are included to visualize pruning variability and tree stability. These plots show the stability of the tree nodes and help determine the optimal size of the tree. This is a crucial decision that balances the risks of overfitting (the tree is too large) and missing moderator interaction effects (the tree is too small).
Learn more about meta-regression analysis and metacart 3.0 in this lightning talk.
If you used AI tools or services to support the preparation of this submission, please state the name and reason for using each of them.
Claude: Grammar check.
Additional Material or Paper
Paper submitted to Statistics in Medicine. Not available yet.
| Keywords: Please list up to 5 keywords to help us find the right session for your contribution. | Meta-analysis, CART, meta-regression, moderator analysis, interaction effects |
|---|---|
| Virtual Option | This submission is for onsite presentation only |
| Material License | CC-BY 4.0 |
| Video Recording | Video sharing is fine |
| The author(s) agree(s) to take responsibility and be accountable for the contents of the submission and is/are authorized to present it. | Confirm |