CBMR: Coordinate-based meta-regression for group and covariate inference

Coordinate-based meta-analysis synthesises findings from multiple neuroimaging studies to identify consistent patterns of brain activation. However, comparing foci distributions between groups of studies remains a challenge, usually requiring balanced sample sizes. In this work, we introduce a multi-group coordinate-based meta-regression (CBMR) framework that employs a generative spline-based spatial model regularised by a roughness penalty, providing flexible control over smoothness. We conduct extensive evaluations with simulations and demonstrate the method on real data. We find that when all groups have at least 200 foci, parametric inference is valid, while sparser datasets require inference via parametric bootstrap. This work is implemented and freely available as a module within the Python package NiMARE, enabling flexible meta-regression and inference for coordinate-based meta-analytic functional MRI datasets involving multiple groups.