Online Program Home
My Program

Abstract Details

Activity Number: 131
Type: Contributed
Date/Time: Monday, August 1, 2016 : 8:30 AM to 10:20 AM
Sponsor: Section on Statistical Learning and Data Science
Abstract #319488 View Presentation
Title: Sparse Spatial Dynamic Factor Model with Basis Expansion
Author(s): Takamitsu Araki* and Shotaro Akaho
Companies: and National Institute of Advanced Industrial Science and Technology
Keywords: Sparse estimation ; Factor loading function ; Basis expansion ; Spatial dynamic factor model

Spatial dynamic factor model (SDFM) represents spatial dependency by a few columns of the factor loading matrix, and also represents temporal behavior by a small number of common factors. The SDFM requires additional estimation of the factor loading on unobserved sites to interpolate the measurements on the sites, and the number of parameters is often large since many the observation sites are needed in scientific research. We propose a sparse SDFM that represents the spatial dependency by factor loading functions, which are smooth continuous functions obtained by a basis expansion, instead of the columns of the factor loading matrix. The basis coefficients are estimated by a sparse regularization method using the extended EM algorithm. The sparse estimation provides more interpretable factor loading function, and can avoid over-parameterization of the model, which leads to stable parameter estimation. The factor loading function enables us to interpolate the measurement at any unobserved point directly. We apply the proposed model and the conventional SDFM to geoscience data in order to compare their performance.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association