Finite mixtures of functional graphical models: Uncovering heterogeneous dependencies in high-dimensional data
2025

Finite Mixtures of Functional Graphical Models

Sample size: 122 publication 10 minutes Evidence: high

Author Information

Author(s): Liu Qihai, Lee Kevin H., Kang Hyun Bin

Primary Institution: Department of Statistics, Western Michigan University, Kalamazoo, MI, United States of America

Hypothesis

Can finite mixtures of functional graphical models effectively capture heterogeneous dependencies in high-dimensional data?

Conclusion

The proposed MFGM method outperforms existing methods in identifying distinct brain connectivity patterns in alcoholics compared to control subjects.

Supporting Evidence

  • The MFGM method effectively identifies heterogeneous subgroups within a population.
  • Simulation studies demonstrate the method's robustness in estimating graphical model parameters.
  • Application to EEG data reveals distinct brain connectivity patterns between alcoholics and controls.

Takeaway

This study created a new way to analyze brain data that helps to see differences between people with alcoholism and those without.

Methodology

The study used a mixture of functional graphical models to analyze EEG data from alcoholics and control subjects, employing an EM algorithm for parameter estimation.

Potential Biases

Potential bias due to the assumption of homogeneity in the data sources.

Limitations

The model assumes that the functional variables follow a multivariate Gaussian process, which may not always hold true.

Participant Demographics

122 subjects, including 77 alcoholics and 45 control subjects.

Digital Object Identifier (DOI)

10.1371/journal.pone.0316458

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication