Pre-conference workshops
Spatio-temporal modelling of areal data
Abstract: This half-day workshop will cover the use of spatio-temporal models for area-level health data, such as case counts in census regions over time. The Besag, York and Mollié model, where relative risk in each region depends on risk in the regions bodering it, with be the foundation of the models presented, and models will be fit using Bayesian inference in R. Various spatio-temporal version of the BYM model will be presented. There will be an emphasis on understanding the computer code used to fit the models and interpreting the results the code produces.
Organizer:
Patrick Brown, University of Toronto and St Michael’s Hospital
Spatial Regression Analysis Using Eigenvector Spatial Filtering
Abstract: This workshop includes hands-on training designed to enable real-world applications of a novel spatial statistical methodology known as Moran eigenvector spatial filtering (MESF) regression, using the recently released ESF Tool, an abridged version of a freely downloadable software package version of SAAR (see Koo et al., 2018, Transactions in GIS). The ESF Tool provides the GUI to a set of integrated functions from open-source software R and DotSpatial, which implements the key workflows for advanced spatial regression modeling with different types of geospatial data and different methods, including the family of spatial autoregressive models for Gaussian random variables (SAR, AR, CAR, SMA, Durbin) and ESF-based models (ESF-OLS, ESF-Poisson, and ESF-binomial). These packages respectively are available at https://thesaar.github.io/ and https://github.com/esftool/esftool. In addition, the workshop will introduce the ESF-based approach to analysis of space-time data and spatial interaction flows, accompanied by R tutorials and demo data sets.
Organizers:
Daniel A. Griffith, Ashbel Smith Professor, The University of Texas at Dallas
Yongwan Chun, The University of Texas at Dallas
Abstract: Maps provide spatially explicit information on many features of the earth’s surface. In particular, environmental monitoring based on information from satellite imagery has evolved to where national, continental, and global maps are routinely produced. Map accuracy assessment is the process of producing descriptive map quality information critical to understanding the potential utility and limitations of a map. This workshop focuses on the methodology to conduct statistically rigorous assessments of map accuracy and the associated objective of estimating area of the categories mapped. The workshop is developed around the three components of accuracy assessment, the response design, sample design, and analysis (i.e., how to determine the “reference” class or value that will be compared to the map, where to collect reference data, and how to use these data to describe accuracy and estimate area). The workshop will emphasize how to decide which methods are best tailored to the objectives and practical constraints of different applications. The primary applications addressed will be “hard” classified (categorical) maps. However, features unique to assessing accuracy of continuous variables will also be addressed. In addition, several special topics that are not addressed by traditional methods of accuracy assessment will be presented. These topics include volunteered geographic information, reference data error, accuracy assessment of object-based maps, and accuracy of time series of maps (e.g., land-cover change). Throughout the workshop case studies of real applications will be used to illustrate the concepts and methods. At the end of the workshop, participants will have: 1) thorough understanding of traditional methods of accuracy assessment; 2) knowledge of the advantages and disadvantages of different choices of sampling design and how to apply that knowledge to choose a design based on project objectives and desirable design criteria; 3) How do decide what sample size to use; 4) how to optimally allocate the sample in a stratified design to meet different accuracy assessment objectives; and 5) awareness of recent new developments in accuracy assessment and area estimation. The content of the workshop is closely related to the methodology presented in Olofsson et al. (2014) and Stehman and Foody (2019).
Organizer:
Dr. Steve Stehman, State University of New York, College of Environmental Science and Forestry