Additional materials for Chapter 9: Linking Spatial Exposure Data to Health Events
(Waller and Gotway (2004) Applied Spatial Statistics for Public Health Data. New York: John Wiley and Sons).
Errata:
p. 332, equation (9.9) The last two terms should be u-squared and v-squared.
p. 363, two lines after equation (9.34). Y-sub-j should be Y(s-sub-j)
p. 371, equation (9.50). x(s-sub-i) should be x(s-sub-j)
Data Sets:
New York leukemia data. (Source: Waller, L.A., Turnbull, B.W., Clark, L.C., and Nasca, P. (1994) "Spatial Pattern Analysis to Detect Rare Disease Clusters" in Case Studies in Biometry, N. Lange, L. Ryan, L. Billard, D. Brillinger, L. Conquest, and J. Greenhouse (eds.) New York: John Wiley and Sons. Waller et al. (1994) originally present data for census block groups in 7 counties, tracts in one county yielding 790 regions. These data are available from StatLib (http://lib.stat.cmu.edu/datasets/csb/, covariate data from 1980 U.S. Census obtained from CensusCD 1980, Version 2.0 by GeoLytics, Inc. (www.geolytics.com)). In Waller and Gotway (2004) we use census tracts for all counties yielding 281 regions (all tracts) in the data set, available below. Initial description in text: pp. 98-104, Chapter 9 Data Breaks: pp. 345-362, 374-379, 423-429, Data listing: pp. 436-442, Waller and Gotway 2004).
Note: In chapter 9 we also use covariate values, so these data sets contain more information than NYTRACT.dat listed in Chapter 4. We also need the adjacency matrix to fit the spatial regression models.
Data set:
Adjacency matrix (a 281 x 281 matrix with 1 in position (i,j) indicating that regions i and j share a common boundary, 0 in position (i,j) indicating no adjacency between regions i and j):
Raccoon rabies data (Data source: U.S. Centers for Disease Control and Prevention, Intitial description in Waller and Gotway (2004): Data Break: pp. 330-333, listing pp. 434-435).
Scottish lip cancer data (Source: Clayton, D.G. and Kaldor, J. (1987) Empirical Bayes estimates of age-standardizated relative risks for use in disease mapping. Biometrics 43, 671-682. Initial description and data listing in Waller and Gotway (2004), Exercise 2.4 (pp. 36-37), Data Break: pp. 392-399).
ESRI shapefile files (this adds two additional variables to allow GIS mapping onto the British National Grid).
R code:
Example of trend surface analysis (coming soon)
Example of linear regression with spatially autocorrelated errors (coming soon)
Examples of spatial autoregressive models (coming soon)
Examples of generalized linear mixed models (coming soon)
SAS code:
Example of linear regression with spatially autocorrelated errors using NY leukemia data (Data breaks on pp. 345-362, 374-379). (nyregress.sas)
Example of generalized linear models with spatially autocorrelated errors using Scottish lip cancer data (Data break on pp. 392-398). (scotglm.sas) Note: this uses the GLIMMIX macro (glmm800.sas), also available as a SAS procedure (http://support.sas.com/news/feature/04sep/statdownloads.html).
(Note: SAS files assume all data is in the folder “c:\myfolder.” Change this specification in the first line to use the code.)
WinBUGS code:
Examples of Bayesian hierarchical models for regional disease data using the NY leukemia data (coming soon)
Note: To run R code,
Note: To run WinBUGS code,