All Times ET
Wednesday, February 2
Wed, Feb 2, 12:30 PM - 1:30 PM
Virtual
Virtual
Poster Session 1
A Practical and Effective Approach for Selecting the Number of Hidden Components (305339)
Haoran Lu, University of Wisconsin, Madison*Chunming Zhang, University of Wisconsin, Madison
Keywords: factor model, feature extraction, imaging data, time series
Choosing the number of hidden components or factors is a practical issue commonly encountered before the data analysis. Though the PCA selection has been frequently used, its drawback has also been known. Inspired by feature extraction and source separation from multi-channel brain EEG recordings and non-linear temporal signal processing, this paper develops a two-step procedure for selecting the number of hidden components, which is broadly applicable in scientific studies. The computational simplicity and practical effectiveness are illustrated with extensive simulation experiments and real data analysis.