Insight Into Individual Differences in Emotion Dynamics With Clustering.

Area of Science:

• Social Sciences
• Psychology
• Computational Social Science

Background:

• Emotion dynamics research increasingly uses time series models.
• Individual emotion dynamics are often heterogeneous, complicating group comparisons.
• Existing dynamic clustering methods assume uniform processes within subgroups.

Purpose of the Study:

• To develop a probabilistic clustering approach for analyzing individual differences in emotion dynamics.
• To overcome the restrictive assumption of equal generating processes in existing methods.
• To enable more accurate comparisons and generalizations of emotion dynamics across diverse groups.

Main Methods:

• A novel probabilistic clustering approach using a mixture model.
• Clustering is based on individuals' vector autoregressive (VAR) coefficients.
• Performance evaluated via simulation and compared to a nonprobabilistic method.

Main Results:

• The proposed probabilistic method demonstrates effective clustering of individuals based on their emotion dynamics.
• Simulation results show comparable or superior performance to nonprobabilistic methods.
• The approach is validated using real-world ecological momentary assessment (EMA) data.

Conclusions:

• Probabilistic clustering on VAR coefficients offers a flexible and powerful tool for studying heterogeneous emotion dynamics.
• This method advances the analysis of individual differences in psychological time series data.
• The findings have implications for understanding and potentially treating conditions like depression and anxiety.