RFX-BMS

This page describes a toolbox for random-effect Bayesian model selection (RFX-BMS) at the group-level.

Here, models are treated as random effects that could differ between subjects, with an unknown population distribution (described in terms of model frequencies/proportions). This differs from a fixed-effect BMS, which assumes that the same model generated the data of all subjects. In brief, RFX-BMS essentially rests on inferring the frequency with which any model prevails in the population.

The toolbox also handles:

• • family inference: large model spaces can be partitionned into model families to boost BMS's statistical power

  • between-conditions BMS: quantifying the evidence for a difference in model attributions across conditions

  • between-groups BMS: quantifying the evidence for a difference in model frequencies across groups

For additional details regarding RFX-BMS, please visit this wiki page.

• Download and install

RFX-BMS is part of a larger toolbox dealing with computational models of behavioural and neurobiological data (VBA).

It requires at least MATLAB 7.1 to run.

You can download the VBA toolbox here.

• Citation

When using the toolbox, please cite the following papers:

Bayesian model selection for group studies - revisited.

L. Rigoux, K. E. Stephan, K. J. Friston, J. Daunizeau

Neuroimage (2014), 84: 971-985 [PubMed]

VBA: a probabilistic treatment of nonlinear models for neurobiological and behavioural data

J. Daunizeau, V. Adam, L. Rigoux

PLoS Comput Biol (2014), 10(1): e1003441 [Pubmed]