Syllabus for NR 245

Advanced Spatial Methods

Spring 2008

image003.png

Instructors: Austin Troy (austin.troy@uvm.edu) and Morgan Grove (jmgrove@gmail.com) 

2 credit—lab course

Room 222 Aiken

Time: Monday 8:45 to 10:45

Austin’s office hours: TBD

 

Description:

This course teaches various statistical and spatial analysis methods through weekly lab exercises and a final project. Among the methods addressed are advanced overlay analysis with geoprocessing and cross-tabulation, Geographically Weighted Regression, spatial cluster analysis, analysis of variance, logistic regression, multi-model inference, spatially weighted regression, analysis of spatial residuals, and measures of spatial autocorrelation. Students will be introduced to S-Plus (including S-Plus spatial module) and GWR software, and will learn new methods in software they have already worked with, including ArcGIS and Microsoft Access. The course currently uses data from the Baltimore Ecosystem Study, an NSF-funded Long Term Ecological Research Project as the focus for all labs. These exercises build sequentially and thematically on a single question from the case study and methods are introduced in the context of answering this question. For instance, the course currently analyzes the relationships between urban green space and socio-economic factors. In each lab, the instructor gives a short lecture on the analytic tools to be covered, giving the statistical, conceptual and mathematical background. Students then apply these concepts in the lab using instructions given on the website.

 

Requirements:

·       Class attendance and participation: 20%

·       Lab exercises: 40%

·       Class project: 40%

 

Class Project: This is a project to be conducted in groups of 2-4 people that focuses on some aspect of spatial analysis, integrated with quantitative or statistical analysis. It should incorporate at least one of the tools that we learned in class, or related tool we did not cover. Students are encouraged to conduct projects related to the Baltimore Ecosystem Study because of the richness and the quality of the dataset (meaning you’ll save time you would have otherwise spent acquiring and processing data). However, if students have other projects they are working on outside of the class, they may use those as the basis for the project if needed data is reasonably available. Students are expected to pose some kind of hypothesis that can be tested using spatial data and statistical methods. Looking at significant differences in variables, relationships, and trends, or developing new approaches for categorizing or segmenting places or populations are all recommended.  The subject can be biophysical, socio-economic or some mix. The scale can range from local (e.g. a city or county), to watershed, to county, to global, although local scale is preferred. The deliverable is a 12-18 page paper (longer for large groups) detailing the research question, a literature review and background, methodology, results and interpretation. It should also include diagrams and maps. Due May 5 at 5

 

Online Readings:

Anselin, L. and A. Getis (1992). Spatial statistical analysis and geographic information systems.  The Annals of regional science 26(1): 19-34.

Anselin, L. and W. K. T. Cho (2000). Spatial Effects and Ecological Inference. Political Analysis 10(3):276-297.

Brunsdon, C, S. Fotheringham, and M. Charlton. 1998. Geographically Weighted Regression-Modeling Spatial Non-stationarity. The Statistician. 47(3):431-443.

Burnham, K and D Anderson. 2002. Model selection and multimodel inference: a practical information theoretic approach. 2nd ed. New York, Springer. Selected pages.

Diniz-Filho, J.A., L.M. Bini and B.A. Hawkins. 2003. Spatial autocorrelation and red herrings in geographical ecology. Global Ecological and Biogeography. (12):53-64.

Fortin, M. and M. Dale. 2005. Spatial Analysis: A guide for ecologists. Cambridge, England: Cambridge University Press.

Goodchild, M. F., L. Anselin, et al. (2000). Toward Spatially Integrated Social Science. International Regional Science Review 23(2): 139-159.

Troy. In Press. Geodemographic segmentation. Encyclopedia of Geographical Information Science

Troy, A. and Grove, M. In Revisions. Property values, parks and crime: a hedonic analysis in Baltimore, MD. Landscape and Urban Planning

Zorn, C. 2003. Agglomerative Clustering of Rankings Data, with an Application to Prison Rodeo Events. Unpublished Working Paper. Emory Univerity.

 

 

  • 1/14: Introduction to the course, GIS refresher, ANOVA
    • Intro and course description
    • Description of BES-LTER and the BES data library
    • Refresher on Analysis of Variance and Box Plots
    • Lab1: Summarizing raster land cover by block group using tabulate areas; model builder; box plots; ANOVA.

 

  • 1/21: MLK day—no class

 

 

  • 2/4:  Local and Global Spatial Autocorrelation
    • Intro to Global metrics of spatial autocorrelation: Moran’s I, Geary’s C
    • Local measures: localmoran, localgeary, LISA
    • Autocorrelation in regression residuals
    • Lab3: look at global and local autocorrelation in variables and residuals
    • Lecture
    • Readings: Diniz-Filho et al; Fortin and Dale, sections 1.2, 3.3 and 3.4 (WebCT)

 

  • 2/11: Intepreting the variogram and basic geostatistics
    • Understanding how variograms describe scales and patterns of autocorrelation and how this information can be used for interpolation
    • Lab4:
      • Plot out empirical variograms in Splus, analyzing variables at several thousand sample points
      • Derive model variograms
      • Try several functional forms and several variables
      • Do ordinary and universal kriging in Arc Map
      • Plot out estimated map and probability map

 

  • 2/18: No class (Presidents’ day)

 

  • 2/25: Linear regression in a spatial context
    • Further analysis of regression diagnostics and assumption tests
    • Lab5: more on regression

 

  • 3/3: Spatially adjusted regression and spatial non-stationarity
    • Intro to spatially adjusted regression
    • Lab6: conducting a spatially adjusted regression and comparing results, including residuals, to regular regression
    • Readings:  Fortin and Dale, sections 5.1.1-5.1.4 (WebCT); Troy and Grove paper (this is an example of regression analysis in action, including spatial regression and Box-Cox transformation)
    • Lecture on spatially adjusted regression approaches

 

 

  • 3/10: NO CLASS (Spring break)

 

  • 3/17: Using GWR to assess nonstationary relationships
    • More on GWR concepts and software
    • Lab7:

·        Use GWR software to run regression between socio-economic variables and tree cover

      • Learn to read GWR output: test local model against global model; test individual parameters for spatial non-stationarity; test gains to model fit
      • Plot out GWR ouput spatially: import map file; spatial join point data to polygons; interpret point-based parameter values, t-statistics and standard errors; plot our spatial residuals of global model.
    • Click here for the GWR manual
    • Click here for a Powerpoint presentation on the use of GWR
    • Reading: Brunsdon, Fotheringham and Charlton (WebCT)

 

  • 3/24: Cluster analysis
    • K Means
    • Partitioning around medoids
    • Readings (for this week and next):
      1. Troy (WebCT)
      2. Zorn (WebCT)
      3. A chapter from Information Theory, Inference and Learning Algorithms by David MacKay (WebCT)

 

  • 3/31: Spatial Cluster Analysis

·        Fuzzy classification, constrained clustering, boundary detection

·        Readings: Fortin and Dale, chapter 4 (Webct)

·        Lab 9

 

 

  • 4/7: Start projects
    • Each group gives brief 5 minute presentation in class of what planning to do.

 

  • 4/14: Projects

 

  • 4/21: Projects

 

  • 4/28: Projects

 

  • Class Project Due Monday May 5 by 5:00