Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Robust Bayesian General Linear Models.

W D Penny1, J Kilner, F Blankenburg

  • 1Wellcome Department of Imaging Neuroscience, University College London, London WC1N 3BG, UK. w.penny@fil.ion.ucl.ac.uk

Neuroimage
|May 8, 2007
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Assessment of differentially methylated loci in individuals with end-stage kidney disease attributed to diabetic kidney disease: an exploratory study.

Clinical epigenetics·2021
Same author

Comparison of methylation patterns generated from genomic and cell-line derived DNA using the Illumina Infinium MethylationEPIC BeadChip array.

BMC research notes·2019
Same author

Erratum to 'Nociceptive activation in spinal cord and brain persists during deep general anaesthesia' (Br J Anaesth 2018; 121: 291-302).

British journal of anaesthesia·2018
Same author

Nociceptive activation in spinal cord and brain persists during deep general anaesthesia.

British journal of anaesthesia·2018
Same author

A validation of dynamic causal modelling for 7T fMRI.

Journal of neuroscience methods·2018
Same author

Event related desynchronisation predicts functional propriospinal myoclonus.

Parkinsonism & related disorders·2016
Same journal

Category-selective neural decreases in the human ventral occipito-temporal cortex as defined with intracranial recordings.

NeuroImage·2026
Same journal

EEG-Based Brain Fingerprints Elicited by Focal Transcranial Magnetic Stimulation of the Primary Motor Cortex.

NeuroImage·2026
Same journal

The Association between Brain Oscillatory Activity and Immediate Memory under Different Magnetoencephalography Paradigms: A population-based Study.

NeuroImage·2026
Same journal

Brain response to awe experiences in virtual reality: an integrated linear and nonlinear EEG analysis.

NeuroImage·2026
Same journal

Convergent imaging and genetic signatures of gray matter atrophy in Parkinson's disease.

NeuroImage·2026
Same journal

What actually matters in multi-compartment EEG head models: A controlled FEM study of parcellation granularity, skull layering, mesh quality, noise, and inverse solver.

NeuroImage·2026
See all related articles

We introduce a Robust General Linear Model (RGLM) using Bayesian learning and a Mixture of Gaussians for noise. This robust regression method enhances coefficient estimation and can adapt to standard models when needed.

Area of Science:

  • Statistics
  • Machine Learning
  • Neuroimaging Analysis

Background:

  • General Linear Models (GLMs) are widely used but sensitive to outliers.
  • Robust regression methods are needed to handle non-Gaussian noise and outliers.
  • Bayesian approaches offer a flexible framework for statistical modeling.

Purpose of the Study:

  • To develop a robust Bayesian learning algorithm for General Linear Models (GLMs).
  • To introduce a novel noise model using a Mixture of Gaussians for enhanced robustness.
  • To enable adaptive model order selection for noise components.

Main Methods:

  • Developed a Bayesian learning algorithm for Robust General Linear Models (RGLMs).
  • Employed a Mixture of Gaussians to model varying noise levels across data points.

Related Experiment Videos

  • Utilized a variational inference framework for regularization and model order selection.
  • Compared the RGLM approach against existing robust regression techniques.

    • Main Results:

    • The proposed RGLM provides robust estimation of regression coefficients by accommodating non-uniform noise.
    • The variational inference framework effectively prevents overfitting and selects the noise model order.
    • The RGLM can automatically default to a standard GLM when robustness is not necessary.
    • Demonstrated the method's efficacy on synthetic datasets and functional Magnetic Resonance Imaging (fMRI) data.

    Conclusions:

    • The Bayesian RGLM with a Mixture of Gaussians offers a powerful and flexible approach to robust regression.
    • This method enhances the reliability of regression coefficient estimation in the presence of outliers or complex noise structures.
    • The algorithm is applicable to various fields, including neuroimaging, where data quality can vary significantly.