Online Program Home
My Program

Abstract Details

Activity Number: 609
Type: Contributed
Date/Time: Wednesday, August 3, 2016 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #320721
Title: Identification of Homogeneous Areas Through Lattice-Based Spatio-Temporal Clustering
Author(s): Rodrigue Ngueyep Tzoumpe* and Huijing Jiang and YoungDeok Hwang
Companies: IBM Research and IBM and IBM T. J. Watson Research Center
Keywords: Time-varying coefficients ; Network lasso ; Convex Clustering

Spatial linear models such as CAR and SAR have been used in several applications to model spatial data collected on a lattice. In most of the existing models spatial dependence is exclusively captured through the covariance structure of the errors.In many spatio-temporal applications a response can depend on a set of predictors, and it is plausible that the relationship of this response to the predictors vary spatially and temporally. An approach relegating the spatio-temporal variation only to the model's error would fail to capture these spatio-temporal changes observed in the relationship of between response and predictors. We propose to explicitly model the spatio-temporal dependence between the response and the predictors in the lattice settings, by assuming that each site on the lattice has its own spatio-temporal relationship. By using fusion penalties between neighboring sites, we jointly estimate the coefficients measuring the impact of explanatory variables on response observed at each site of the lattice. We apply to proposed method to a simulated data sets and to county level spatio temporal data.

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

Back to the full JSM 2016 program

Copyright © American Statistical Association