bayesian-ippm-entrainment-evidence

name	bayesian-ippm-entrainment-evidence
description	Bayesian framework for quantifying neural entrainment evidence in Information Processing Pathway Maps (IPPMs). Replaces frequentist null hypothesis testing with probabilistic model adjudication using Bayes factors. Enables robust comparison of competing computational models explaining neural data, with explicit handling of collinear models.
activation	Bayesian entrainment, IPPM, neural entrainment evidence, model adjudication, Bayes factor, auditory processing pathway, cortical entrainment, frequentist vs Bayesian
tags	["neuroscience","bayesian-inference","neural-entrainment","model-comparison","auditory-processing","EEG","MEG"]
version	1.0.0
author	agent
arxiv_id	2607.06284
paper_title	Quantifying Entrainment Evidence: A Comparison of Frequentist and Bayesian Approaches for Information Processing Pathway Maps

Bayesian Framework for Neural Entrainment Evidence (IPPMs)

Core Innovation

Problem

Information Processing Pathway Maps (IPPMs) formalize the sequence of mathematical transformations applied to sensory stimuli, mapping latency and cortical expression of computational steps. Traditionally relies on frequentist hypothesis testing to link model outputs with observed neural activity.

Limitation: Determining which of several competing computational models best explains neural data is a problem of model adjudication, not null hypothesis rejection.

Solution

Bayesian framework that:

• Retains IPPM core strength: generating explicit predictions of time-varying neural signals
• Shifts selection criterion from rejecting null to quantifying relative evidence for competing hypotheses
• Handles collinear models more robustly

Methodology

IPPM Framework

Sensory Stimulus -> [Computational Model 1] -> Predicted Neural Signal
Sensory Stimulus -> [Computamental Model 2] -> Predicted Neural Signal
                                    |
                        Compare with Observed Neural Activity
                                    |
                        Model Adjudication (Bayesian)

Bayesian Formulation

1. Prior: Define priors over competing computational models
2. Likelihood: Compute likelihood of observed neural data given each model's predictions
3. Posterior: Update model probabilities using Bayes' rule
4. Bayes Factor: Quantify relative evidence between models

Key Mathematical Components

• Time-varying neural signal predictions from each computational model
• Bayesian model comparison using marginal likelihoods
• Evidence accumulation across subjects/conditions

Comparison: Frequentist vs. Bayesian

Frequentist Approach (Traditional)

• Null hypothesis: Model predictions differ from observed data
• Test: Reject null if p < threshold
• Limitation: Cannot quantify evidence FOR a model, only against null
• Problem: Multiple competing models lead to multiple tests and inflated false positive rate

Bayesian Approach (Novel)