19 July 2023
Proposes robust Bayesian FPCA using skew elliptical distributions
Handles asymmetry and outliers better than conventional methods
Models covariance function via spectral decomposition with priors
Extends to sparse functional data unlike previous Bayesian FPCA
Simulation studies demonstrate superior performance with outliers
Capturing variation in functional data via Bayesian modeling
This paper develops a robust Bayesian approach for functional principal component analysis. It models observations using skew elliptical distributions to handle asymmetry and outliers. The method approximates the covariance function via spectral decomposition and places priors on parameters. It extends the model to sparse functional data. Simulation studies show the method outperforms others with outliers. Applications to environmental and biological data illustrate effectiveness in identifying abnormal curves.
No comments yet, be the first to start the conversation...
Sign up to comment on this paper