Regularized Multilevel Multinomial Regression for Select-All-That-Apply Responses and High-Dimensional Predictors with Applications to Perception of Facial Expressions
Analytical framework connects mouth shape variations to perceived emotions in facial expressions, suggesting new insights for understanding human emotions.
Key Points
To develop a method for quantifying mouth shape variation and its relationship to perceived emotions from facial expressions.
Analyzed ratings from 802 individuals on 27 smile-like expressions using open-source data.
Utilized statistical shape analysis with 30 landmarks to parameterize mouth shapes.
Employed a nonparametric multinomial regression model for high-dimensional predictors to relate mouth shape features to emotion ratings.
The three-dimensional representation of landmark coordinates achieved better predictive performance than full-dimensional sets.
Easily interpretable predictions were produced, enhancing understanding of mouth shape variations' impact on emotional perception.