Related Experiment Videos
The structure of self-reported emotional experiences: a mixed-effects Poisson factor model
Ulf Böckenholt1, Wagner A Kamakura, Michel Wedel
1University of Groningen, The Netherlands. ulf.bockenholt@mcgill.ca
The British Journal of Mathematical and Statistical Psychology
|November 25, 2003
Summary
Analyzing emotion data, a standard Poisson model proved insufficient. A novel factor-analytic Poisson model offers a computationally efficient and accurate solution for understanding emotion experiences.
Area of Science:
- Statistics
- Psychology
- Computational Modeling
Background:
- Multivariate count data analysis often employs Poisson distributions with random effects.
- Standard single random-effect Poisson models may not adequately fit complex datasets like emotion frequencies.
- High-dimensional integration and numerous parameters pose challenges for alternative models.
Purpose of the Study:
- To address the limitations of standard Poisson models for multivariate emotion count data.
- To propose a computationally feasible and statistically robust alternative.
- To introduce a factor-analytic Poisson model for analyzing emotion experiences.
Main Methods:
- Development and application of a factor-analytic Poisson model.
- Utilizing marginal maximum likelihood for parameter estimation.
- Incorporating goodness-of-fit tests and covariate analysis.
Main Results:
- A two-dimensional factor-analytic Poisson model provided an excellent fit to the emotion data.
- The model identified two key factors: pleasantness/unpleasantness and activation levels of emotions.
- The proposed model overcomes computational challenges associated with high-dimensional random-effects models.
Conclusions:
- The factor-analytic Poisson model is a powerful tool for analyzing multivariate emotion count data.
- This approach offers a parsimonious and interpretable solution for understanding emotional structures.
- The model facilitates rigorous testing of random-effects structures and the incorporation of covariates.