Bayesian Inference Will Penny Wellcome Department of Imaging Neuroscience, University College London, UK SPM Course, London, May 2004
Bayesian Inference Gaussians Posterior Probability Maps (PPMs) Hemodynamic Models (HDMs) Comparing models (DCMs)
One parameter Likelihood and Prior Posterior Likelihood Prior Relative Precision Weighting
Two parameters
Posterior Probability Maps - PPMs Bayesian parameter estimation Least squares parameter estimation No Priors Shrinkage priors
Prior Prior: mi 1/a=1
Data Likelihood: yi i=1..V=20 voxels n=1..N=10 scans
Least Squares Estimates
Least Squares Estimates Error=7.10
Bayesian Estimates E-Step: M-Step: Iteration 1 Error=7.10
Bayesian Estimates E-Step: M-Step: Iteration 2 Error=2.95
Bayesian Estimates E-Step: M-Step: Iteration 3 Error=2.35
SPMs and PPMs PPMs: Show activations of a given size SPMs: show voxels with non-zero activations
Hemodynamic Models - HDMs Photic Motion Attention fMRI study of attention to visual motion
PPMs Advantages Disadvantages Use of shrinkage One can infer a cause priors over voxels is computationally demanding Utility of Bayesian approach is yet to be established One can infer a cause DID NOT elicit a response SPMs conflate effect-size and effect-variability
Comparing Dynamic Causal Models V1 V5 SPC Motion Photic Attention 0.86 0.56 -0.02 1.42 0.55 0.75 0.89 V1 V5 SPC Motion Photic Attention 0.96 0.39 0.06 0.58 m=3 V1 V5 SPC Motion Photic Attention 0.85 0.57 -0.02 1.36 0.70 0.84 0.23 Bayesian Evidence: Bayes factors: