Factor analysis linking functions for simultaneously modeling neural and behavioral data.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Neuroscience

Background:

• Integrating cognitive models with neurophysiology is crucial for advancing cognitive neuroscience.
• Current joint modeling frameworks face challenges due to the high dimensionality of neural data.
• A need exists for methods that can effectively link abstract cognitive models with complex neural recordings.

Purpose of the Study:

• To introduce a novel linking function based on factor analysis for joint modeling in cognitive neuroscience.
• To address the limitations imposed by high-dimensional neural data in current linking methods.
• To enhance the scalability and accuracy of integrating cognitive and neural data.

Main Methods:

• Developed a new linking function utilizing factor analysis, allowing linear complexity growth with neural features.
• Evaluated the linking function through two simulation studies, assessing parameter recovery and model performance.
• Applied the new linking function to real-world data from a perceptual decision-making task.

Main Results:

• The new linking function accurately recovers model parameters even with a large number of neural features.
• It successfully reconstructs the data-generating model, even under model misspecification.
• The proposed function outperforms previous methods in cross-validation tests.
• Analysis of perceptual decision-making data revealed how brain function differences relate to model parameters under speed and accuracy instructions.

Conclusions:

• The proposed factor analysis-based linking function offers a scalable and robust solution for joint cognitive and neural modeling.
• This advancement facilitates a more nuanced understanding of the relationship between cognitive processes and underlying neural mechanisms.
• The method has practical implications for analyzing complex neural data in cognitive neuroscience and related fields.