Explore a comprehensive lecture on large-scale high temporal resolution effective connectivity analysis in MEG and the challenges of inference. Delve into the complexities of analyzing multiple influencing factors in judgments and the limitations of post-interaction observations. Examine BOLD imaging's potential and challenges in identifying key processing components and localizing them based on empirical literature. Learn about Granger Causation and its implementation through lagged vector autoregression models, as well as the critical assumptions and requirements of classical Granger causation. Discover the GPS method for implementing Granger's assumptions with integrity, including imaging considerations and ROI identification techniques. Investigate the stationarity problem in prediction and the application of Kalman filters. Study data reduction through graph theory and the comparison of experimental conditions. Gain insights into neural decoding techniques using the same data to probe representation. Conclude by addressing the challenges and opportunities in this field of neuroscience research.
Large Scale High Temporal Resolution Effective Connectivity Analysis
Overview
Syllabus
Intro
Outline
Challenge: When Multiple Factors Influence Judgements
Two Explanations of the Same Results
Task effects, decision mechanisms, response blases...
and the problem of making observations after the interactions have been completed
BOLD Imaging: Promise and Challenges
Identify key components of processing models
Localize components based on empirical literature
Determine Pattern of Effective Connectivity
Two intuitions about cause and effect
Granger Causation: Implementing Wiener's definition of causality
Implementation: Prediction by lagged vector autoregression model (VAR)
Critical assumptions requirements of classical Granger causation
GPS: Implementing Granger's Assumptions with Integrity
Imaging Considerations
Identifying Rols: 3. Eliminate redundant ROis based on timeseries comparison
Identifying Rols: 3. Define ROI around centroid based on timeseries comparison
Prediction and the Stationarity Problem
Kalman Filter: Model, Predict, Evaluate, Update
Measuring Granger Causation
Data Reduction Through Graph Theory
Afferent/Efferent Relationship between two
Comparison between experimental conditions
GPS: Our processing stream to automate the Granger Analysis of MR-constrained MEG/EEG data
Neural Decoding: Using the same data to probe representation
Challenges and Opportunities
Taught by
MITCBMM
