Modeling continuous self-report measures of perceived emotion using generalized additive mixed models

Area of Science:

• Psychology
• Cognitive Science
• Affective Science

Background:

• Traditional emotion research relies on static facial expressions, limiting the study of dynamic emotional responses.
• Continuous self-report methods capture dynamic emotion, but lack standardized statistical analysis techniques.
• Existing methods struggle to account for the temporal and individual variations inherent in dynamic emotion data.

Purpose of the Study:

• To introduce and validate advanced statistical models for analyzing dynamic emotion expressions.
• To provide a consensus on appropriate statistical techniques for continuous self-report emotion data.
• To enable robust inferences about group differences in dynamic emotion responses.

Main Methods:

• Application of generalized additive models (GAMs) and generalized additive mixed models (GAMMs) to continuous emotion expression data.
• Modeling shared components of perceived emotion across time while accounting for individual differences.
• Utilizing GAMMs to address autocorrelation in time series data and model participants as random effects.

Main Results:

• GAMs and GAMMs effectively analyze the dynamic nature of continuous emotion expression measures.
• The generalized additive mixed model (GAMM) approach is superior due to its ability to handle autocorrelation and random effects.
• Proposed methods enhance confidence in detecting linear differences between groups in dynamic emotion responses.

Conclusions:

• Generalized additive models (GAMs) and generalized additive mixed models (GAMMs) offer powerful tools for analyzing dynamic emotion expressions.
• GAMMs provide a statistically rigorous framework for understanding temporal and individual variations in emotion.
• These advanced statistical techniques advance the field of emotion research by enabling more nuanced and accurate analyses.