From observed laterality to latent hemispheric differences: Revisiting the inference problem.

Area of Science:

• Neuroscience
• Cognitive Psychology
• Biostatistics

Background:

• Inferring latent functional brain differences from observed lateralization is common in hemispheric dominance research.
• Current methods often neglect antecedent probabilities, leading to misclassifications, particularly for low-prevalence groups.

Purpose of the Study:

• To present a general Bayesian framework for valid inference of hemispheric dominance.
• To apply this framework to infer speech processing dominance from dichotic listening scores.
• To provide tools for exploring the boundaries of valid inference.

Main Methods:

• Development of a general Bayesian inferential framework.
• Application to dichotic listening data for speech processing lateralization.
• Utilizing large-scale empirical datasets and simulation studies.
• Implementation in R package and Shiny app.

Main Results:

• The proposed Bayesian framework enables valid inferences of hemispheric dominance, including for low-probability classes.
• Inference accuracy is bounded, necessitating careful consideration of prior probabilities.
• Repeated laterality measures can enhance classification outcomes.

Conclusions:

• A robust Bayesian approach is crucial for accurate hemispheric dominance inference.
• The framework and associated tools facilitate more reliable classification of brain lateralization.
• This methodology addresses limitations in current research practices regarding antecedent probabilities.