Comparing Dynamic Causal Models using AIC, BIC and Free Energy
By W. D. Penny. NeuroImage, 2012
311 citations recorded by the source corpus.
In neuroimaging it is now becoming standard practise to fit multiple models to data and compare them using a model selection criterion. This is especially prevalent in the analysis of brain connectivity. This paper describes a simulation study which compares the relative merits of three model selection criteria (i) Akaike's Information Criterion (AIC), (ii) the Bayesian Information Criterion (BIC) and (iii) the variational Free Energy. Differences in performance are examined in the context of General Linear Models (GLMs) and Dynamic Causal Models (DCMs). We find that the Free Energy has the best model selection ability and recommend it be used for comparison of DCMs.
Subfields: Artificial Intelligence, Statistics and Probability
View Paper: https://www.semanticscholar.org/paper/cd839376c8b166be2b3014ced98a8580f1b49a90
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- ChrisD . Frith
- Anil K. Seth
- Michael Breakspear
- Vladimir Litvak
- Giovanni Pezzulo
- Cathy J. Price
- Karl J. Friston
- Rosalyn J. Moran
In Statistics and Probability
- Karl J. Friston
- Vladimir Litvak
- Mohamed L. Seghier
- Peter Zeidman
- Cathy J. Price
- ChrisD . Frith
- Amirhossein Jafarian
- Waldemar Karwowski
Citations in the corpus
- A guide to group effective connectivity analysis, part 2: Second level analysis with PEB (2019, methodology): “Bayesian Model Comparison and Reduction Hypotheses are tested in PEB by comparing the evidence for different models, which is called Bayesian model comparison or selection (c.f. Penny, 2012).”
- Application of Graph Theory for Identifying Connectivity Patterns in Human Brain Networks: A Systematic Review (2019, methodology): “Model selection in DCM is based on a direct comparison between all models (Penny, 2012), whereas in GC this involves testing for the presence of GC followed by selecting the VAR model order using Akaike or Bayesian information criteria (Bressler and Seth, 2011).”
- Active Inference, Curiosity and Insight (2017, background): “In other words, strong evidence for the reduced model relative to a full model means that the reduced model is about 20 times more likely (Penny, 2012).”
- Criticality in the brain: A synthesis of neurobiology, models and cognition. (2017, methodology): “upon data analysis. Third, computational models of neural systems that are based upon criticality should be tested closely against empirical data using appropriate frameworks (Daunizeau et al., 2009; Penny, 2012) that allow disambiguation against competing models that invoke other nonlinear mechanis”
- Analysing connectivity with Granger causality and dynamic causal modelling (2013, methodology): “Dynamic causal modelling is introduced as the Bayesian inversion of dynamic (bilinear) neurophysiological models of fMRI time-series [2] Bayesian model comparison is described for selecting among alternative DCMs [38] Neural drivers in DCM for fMRI validated using concurrent electrophysiology in rod”
- Bayesian model selection for group studies - Revisited (2014, unknown): “This replicates the results in Penny (2012).”