next up previous index
Next: asa.gen.meth.02 Up: ASA General Methodology (2) Previous: ASA General Methodology (2)

asa.gen.meth.01


Sponsoring Section/Society: General Methodology

Session Slot: 8:30-10:20 Wednesday

Estimated Audience Size: xx-xxx

AudioVisual Request: xxx


Session Title: Mapping: Statistical Challenges for Inference and Visualization

Theme Session: Yes/No

Applied Session: Yes/No


Session Organizer: Sedransk, Nell Case Western Reserve University


Address: Department of Statistics Case Western Reserve University 339 Yost Hall Cleveland, Ohio 44106-7054

Phone: 216-368-6941

Fax: 216-368-0252

Email: nxs20@po.cwru.edu or lfs2@po.cwru.edu


Session Timing: 110 minutes total (Sorry about format):

Opening Remarks by Chair - 0 minutes First Speaker - 25 minutes Second Speaker - 25minutes Third Speaker - 25 minutes Fourth Speaker- 25 minutes Floor Discussion - 10 minutes


Session Chair: TBN


Address:

Phone:

Fax:

Email:


1. Flexibility in Hierarchical Models of Disease Rates

Waller, Lance A.,   University of Minnesota


Address: Division of Biostatistics Box 303 Mayo Building 420 Delaware Street SE University of Minnesota Minneapolis, Minnesota 55455-0592

Phone: 612-625-2138

Fax: 612-626-0660

Email: lance@muskie.biostat.umn.edu

Best, Nicky, Imperial College of Medicine at St. Mary's

Abstract: We consider the impact of different implementations of hierarchical models for disease mapping. We model regional disease counts as Poisson random variables, incorporating covariate information in the log-relative risk. Prior distributions of random effects allow excess heterogeneity and spatial smoothing of rates toward neighboring values. Using simulated data, we assess the impact of prior and hyperprior specifications, and of spatial neighbor definitions. Of particular interest is the accuracy of spatial smoothing in recovering disease patterns due to spatially structured covariates omitted from the model.


2. Hierarchical Spatial-Temporal Modeling with Geophysical Applications

Berliner, L. Mark,   Ohio State University


Address: Department of Statistics Ohio State University 1958 Neil Ave. Columbus, OH 43210-1247

Phone: 614-292-0291

Fax: 614-292-2096

Email: mb@stat.ohio-state.edu

Royle, J. Andy, National Center for Atmospheric Research

Wikle, Christopher K., National Center for Atmospheric Research

Milliff, Ralph, National Center for Atmospheric Research

Abstract: The role of Bayesian hierarchical modeling of spatial-temporal geophysical phenomena is reviewed. An illustrative example, primarily focusing on spatial structure, is presented. In this example a multivariate hierarchical model is developed for wind and pressure fields over the Labrador Sea, based on satellite data. This model is stochastic, but makes use of dynamical relationships within the probability framework. A key feature of this approach is that the relationship between the gradient of pressure and wind components provides the dynamical basis for a stochastic model of the wind, conditional on the pressure field. Furthermore, variations from this gradient relationship are possible due to the stochastic nature of the formulation. Although the methodology can be used without pressure observations, such observations can be included. Extensions to temporal models are discussed.


3. Nonparametric Regression for Geographic Visualization and Analysis of Environmental Policy

Whittaker, Gerald,   Economic Research Service, USDA


Address: United States Department of Agriculture Economic Research Service USDA 1800 M Street Room 4177 Washington, DC 20036-5831

Phone: 202-694-5557

Fax: 202-694-5775

Email: gerryw@econ.ag.gov

Scott, David W., Rice University

Abstract: The U.S. Department of Agriculture administers and analyzes a large number of surveys of both economic and environmental data. The data collected have traditionally been analyzed on the basis of large geographic regions, and results presented in tables. Through the use of the averaged shifted histogram (ASH) approach to nonparametric regression, it is now possible to analyze survey data on the basis of local areas. The results can be presented as surfaces, or maps where the data are geo-referenced. The spatial analysis of covariates using the ASH has provided new insights into policy applications such as the relation of government payments to agricultural land values, spatial distribution of government payments, and use of agricultural nutrients. In a related use of the ASH estimator, we use estimates of surfaces from survey data to link economic and physical models for analysis of environmental issues. Much economic and environmental data are georeferenced to points which are drawn from a spatial distribution. With the ASH estimator data sets which are georeferenced to different sets of points can be linked for analysis. ASH estimates of surfaces representing physical variables allow analysis that links firm-level economic decisions to spatially distributed processes in the physical environment. The economic and physical effects of alternative policies can then be modeled and the results visualized with nonparametric regression. The ASH is very fast compared to most smoothers, and can be modified to account for a complex survey design in estimation. In this paper, the basic ASH methodology is described and several case studies presented.


4. Recent Developments in Spatial Graphics From Micromaps to Global Grids

Carr, Daniel B.,   George Mason University


Address: George Mason University Department of Applied and Engineering Statistics George Mason University 9930 Rand Drive Burke, Virgina 22015-3812

Phone: 703-993-1671

Fax: 703-993-1700

Email: dcarr@galaxy.gmu.edu

Olsen, Anthony R., US EPA, Corvallis, OR

Courbois, Jean-Yves (Pip) , Oregon State University, Corvallis, OR

Pierson, Suzanne, OAO Corporation, Corvallis, OR

Kimerling, Jon, Oregon State University

Sahr, Kevin, University of Oregon

White, Denis, US EPA, Corvallis, OR

Abstract: In this talk we focus attention two recent methodological developments for spatial graphics, linked micromap plots (LM plots) and icosahedral Synder equal area (ISEA) global grids. LM plots follow a new template for showing spatially indexed statistics more accurately than traditional choropleth maps. A LM plot consists of parallel sequences of micromap, label and statistical summary panels. The micromaps are often low resolution caricatures that show regions or sites with just enough detail to preserve region recognition and neighbor relationships. The statistical summary panels often take familiar forms such as dot plots, bar plots and box plots, time-series plots and scatterplots. Sorting plus logical or perceptual grouping defines the sequence of panels. Color and position link corresponding graphics elements for each case within a set of panels. The careful design makes LM plots a good device for communicating spatially indexed statistical summaries. Some examples address data sets with millions of observations.

In the last part of the talk, we call attention to ISEA global grids of Jon Kimerly and his colleagues Denis White and Kevin Sahr. The ISEA global grids provide a coherent framework for addressing recurrent issues in environmental sampling and have strong implications for spatial data storage, manipulation, analysis, and presentation. The best known global gridding framework is an equal angular grid that is closely related to the Mercator projection and it's well-known polar distortions. ISEA grids provide global uniformity in terms of cell area. When projected onto an icosahedron, the cells are regular hexagons. While issues remain in regard to changing resolution, ISEA grids may eventually emerge as the preferred framework for integrating data from multiple sources and multiple disciplines.

List of speakers who are nonmembers: None


next up previous index
Next: asa.gen.meth.02 Up: ASA General Methodology (2) Previous: ASA General Methodology (2)
David Scott
6/1/1998