ESE 502 Home Page




Instructor: Tony E. Smith


274 Towne (898-9647)
tesmith@seas.upenn.edu


Office Hours:
By Appointment

     Phil_igc_image

 

 


 

TABLE OF CONTENTS



COURSE DESCRIPTION

The course is designed to introduce students to modern statistical methods for analyzing spatial data. These methods include nearest-neighbor analyses of spatial point patterns, variogram and kriging analyses of continuous spatial data, and autoregression analyses of areal data. The underlying statistical theory of each method is developed and illustrated in terms of selected GIS applications. Students are also given some experience with ARCMAP, JMP, and MATLAB software.
 

(return to contents)


PREREQUISITES

  ESE 302 or MUSA 501 (or comparable course in statistics and/or regression)

(return to contents)


REQUIRED MATERIALS

Required Text

Interactive Spatial Data Analysis, T.C. Bailey and A.C. Gatrell

Also Recommended:


(return to contents)


COURSE TOPICS

Spatial Point Pattern Analysis

Continuous Spatial Data Analysis

Regional Data Analysis


COURSE GRADING

There are no exams in this course. Grading is based entirely on the seven homework assignments. Each assignment is weighted equally in the final evaluation. But in "border-line" cases for final grades, I do look for improvement over the semester.

(return to contents)


COURSE TIME AND LOCATION 
 


(return to contents)


CETS Labs
 


(return to contents)


TENTATIVE SCHEDULE FOR SPRING 2014


Lectures Day/Date Topic Homework
INTRO Th/Jan.16 Introduction
1 Tu/Jan.21 Point Pattern Data
2 Th/Jan.23 CSR Hypothesis
3 Tu/Jan.28 Nearest-Neighbor Methods
4 Th/Jan.30 Data Applications PS1 due
5 Tu/Feb.4 K-Function Analysis
6 Th/Feb.6 Simulation Testing Methods
7 Tu/Feb.11 Bivariate K-Functions
8 Th/Feb.13 Tests of Pattern Similarity
9 Tu/Feb.18 Local K-Functions  
10 Th/Feb.20 Continuous Spatial Data
11 Tu/Feb.25 Spatial Variograms PS2 due
12 Th/Feb.27 Variogram Estimation
13 Tu./Mar.4 Simple Kriging Model
14 Th/Mar.6 Kriging Predictions  
  Tu/Mar.11 SPRING BREAK

Th/Mar.13 SPRING BREAK

15
Tu/Mar.18 Simple Regression Model PS3 due
16
Th/Mar.20 Generalized Least Squares
17
Tu/Mar.25 Universal Kriging Model
18 Th/Mar.27 Universal Kriging Estimation  
19 Tu/Apr.1 Data Applications  
20 Th/Apr.3 Data Applications PS4 due
21 Tu/Apr.8 Regional Spatial Data
22 Th/Apr.10 Spatial Autocorrelation
23 Tu/Apr.15 Spatial Concentration  
24 Th/Apr.17 Spatial Autoregression PS5 due
25 Tu/Apr.22 Spatial Lag Model 
26 Th/Apr.24 Spatial Diagnostics
27 Tu/Apr.29 Additional Regression Topics PS6 due
PS7 Mon/May 5 Last Assignment PS7 due

(return to contents)


HOMEWORK ASSIGNMENTS

When doing the lab procedures for these assignments, you might wish to work together with other class members. This is often a more efficient way to learn, and is certainly more fun. My only requirement is that your written reports be individual work.

When submitting assignments, I would like you to email me a PDF copy for my files and submit a HARD copy for grading purposes. You can submit your assignments in class, or leave them in the box on the floor outside my office door (by 5PM on the due date).

Example Assignment
Example Answer
Assignment 1
Assignment 2
 Assignment 3
 Assignment 4
 Assignment 5
 Assignment 6
Assignment 7

(return to contents)
 


EXTRA MATERIALS

Using LeSage Models.     This gives a brief introduction to the package of MATLAB
                                        programs available on Jim Lesage's web site.


ArcView 10 Manual.         This manual, written by Amy Hillier, provides a brief
                                         introduction to some of the more useful procedures
                                         in ArcView 10. 


Using MATHTYPE.            These notes show you how to access MATHTYPE in
                                         WORD, and use it to write both mathematical equations
                                         and in-line expresssions in your reports.

Matrix Regression.              These slides give an introduction to MATRIX ALGEBRA
                                         in the context of  MULTILE REGRESSION


CML Data and Maps.        This is a set of instructions on how to download data sets
                                         and map shapefiles from the Neighborhood Information
                                         System (NIS) in the Cartographic Modeling Lab (CML).

Reference Materials.           This is a secure site containing additional reference materials
                                         for the course that are copyrighted.


Lab Data Access.                 These notes are intended for those students who do not have

                                         a class account, but would still like to access the class data sets and

                                         use the software in the lab.                                      


Remote Data Access.           These notes are intended for those students who have access to
                                          the ARCMAP, JMP, and MATLAB software elsewhere, and
                                          want to access the class data sets from this remote location.
                                         

 

(return to contents)
    


SPATIAL DATA WEB SITES


PENN CAMPUS RESOURCES

http://guides.library.upenn.edu/data
http://www.cml.upenn.edu/
 

GENERAL SPATIAL DATA RESOURCES

http://www.census.gov/geo/www/cob/index.html

http://www.dcrp.ced.berkeley.edu/research/footprint
http://fisher.lib.virginia.edu/index.html
http://www.gisdatadepot.com/

http://gis.about.com/
http://www.esri.com/data/download/index.html
http://www.maproom.psu.edu/dcw/
http://www.lib.ncsu.edu/stacks/gis/dcw.html
http://www.fedstats.gov/mapstats/
http://factfinder.census.gov/servlet/BasicFactsServlet
http://homer.ssd.census.gov/cdrom/lookup
http://www.nationalgeographic.com/maps/
http://nationalatlas.gov/natlas/natlasstart.asp
http://www3.cancer.gov/atlasplus/
http://www.wakanow.com/ng/pages/a-wakanow-guide-to-geography

(return to contents)
 


NOTEBOOK ON SPATIAL DATA ANALYSIS

INTRODUCTION 

I. SPATIAL POINT PATTERN ANALYSIS

    1. Examples of Point Patterns
        1.1  Clustering versus Uniformity
        1.2  Comparisons between Point Patterns

    2. Complete Spatial Randomness
 
      2.1  Spatial Laplace Principle

        2.2  Complete Spatial Randomness
       
2.3  Poisson Approximation
        2.4  Generalized Spatial Randomness
        2.5  Spatial Stationarity
   
    3. Testing Spatial Randomness

        3.1  Quadrat Method
        3.2  Nearest-Neighbor Methods
               3.2.1  Nearest-Neighbor Distribution under CSR
               3.2.2  Clark-Evens Test
        3.3  Redwood Seedling Example
               3.3.1  Analysis of Redwood Seedlings using JMPIN
               3.3.2  Analysis of Redwood Seedlings using MATLAB
        3.4  Bodmin Tors Example
        3.5  A Direct Monte Carlo Test of CSR

    4.  K-Function Analysis of Point Patterns

       
4.1  Wolf-Pack Example
        4.2  K-Function  Representations
        4.3  Estimation of K-Functions
        4.4  Testing the CSR Hypothesis
        4.5  Bodmin Tors Example
        4.6  Monte Carlo Testing Procedures
               4.6.1  Simulation Envelopes
               4.6.2  Full P-Value Approach
        4.7  Nonhomogeneous CSR Hypotheses
               4.7.1  Housing Abandonment Example
               4.7.2  Monte Carlo Tests of Hypotheses
               4.7.3  Lung Cancer Example
         4.8  Nonhomogeneous CSR Hypotheses
               4.8.1  Construction of Local K-Functions
               4.8.2  Local Tests of Homogeneous CSR Hypotheses
               4.8.3  Local Tests of Nonhomogeneous CSR Hypotheses

    5.  Comparative Analyses of Point Patterns

        
5.1  Forest Example
         5.2  Cross K-Functions
         5.3  Estimation of Cross K-Functions
         5.4  Spatial Independence Hypothesis
         5.5  Random-Shift Approach to Spatial Independence
               5.5.1  Spatial Independence Hypothesis for Random Shifts
               5.5.2  Problem of Edge Effects
               5.5.3  Random Shift Test
               5.5.4  Application to the Forest Example
         5.6  Random-Labeling Approach to Spatial Independence
               5.6.1  Spatial Indistinguishability Hypothesis
               5.6.2  Random Labeling Test
               5.6 3  Application to the Forest Example
         5.7  Analysis of Spatial Similarity
               5.7.1  Spatial Similarity Test
               5.7.2  Application to the Forest Example
         5.8  Larynx and Lung Cancer Example
               5.8.1  Overall Comparison of the Larynx and Lung Cancer Populations
               5.8.2  Local Comparison in the Vacinity of the Incinerator
               5.8.3  Local Cluster Analysis of Larynx Cases

    6.  Space-Time Point Processes

        
6.1  Space-Time Clustering
         6.2  Space-Time K-Functions
         6.3  Temporal Indistinguishability Hypothesis
         6.4  Random Labeling Test
         6.5  Application to the Lymphoma Example

      APPENDIX TO PART I
 

II. CONTINUOUS SPATIAL DATA ANALYSIS

  1. Overview of Spatial Stochastic Processes


        1.1  Standard Notation
        1.2  Basic Modeling Framework

  2. Examples of Continuous Spatial Data


        2.1  Rainfall in the Sudan

        2.2  Spatial Concentration of PCBs

  3. Spatially-Dependent Random Effects


        3.1  Random Effects at a Single Location    

               3.1.1   Standardized Random Variables
               3.1.2   Normal Distribution

               3.1.3   Central Limit Theorems

               3.1.4   CLT for the Sample Mean
        3.2  Multi-Location Random Effects  

               3.2.1   Multivariate Normal Distribution

               3.2.2   Linear Invariance Property

               3.2.3  Multivariate Central Limit Theorem

        3.3  Spatial Stationarity  

               3.3.1   Example: Measuring Ocean Depths

               3.3.2   Covariance Stationarity

               3.3.3   Covariograms and Correlograms

  
  4. Variograms


        4.1  Expected Squared Differences

        4.2  The Standard Model of Spatial Dependence

        4.3  Non-Standard Spatial Dependence   
        4.4  Pure Spatial Dependence

        4.5  The Combined Model   

        4.6  Explicit Models of Variograms   
              4.6.1  The Spherical Model

              4.6.2  The Exponential Model   

              4.6.3  The Wave Model  
        4.7  Fitting Variogram Models to Data    
              4.7.1  Empirical Variograms

              4.7.2  Least-Squares Fitting Procedure

        4.8  The Constant-Mean Model

        4.9  Example: Nickel Deposits on Vanvouver Island   
              4.9.1  Empirical Variogram Estimation

              4.9.2  Fitting a Spherical Variogram

      4.10  Variograms versus Covariograms
               4.10.1  Biasedness of the Standard Covariance Estimator

               4.10.2  Unbiasedness of Empirical Variogram for Exact-Distance Samples

               4.10.3  Approximate Unbiasedness of General Empirical Variograms

  
  5. Spatial Interpolation Models


        5.1  A Simple Example of Spatial Interpolation

          5.2  Kernel Smoothing Models

        5.3  Local Polynomial Models

        5.4  Radial Basis Function Models

        5.5  Spline Models

        5.6  A Comparison of Models using the Nickel Data

6. Simple Spatial Prediction Models


        6.1  An Overview of Kriging Models

                   6.1.1  Best Linear Unbiased Predictors

               6.1.2  Model Comparisons

          6.2  The Simple Kriging Model

                   6.2.1  Simple Kriging with One Predictor

                   6.2.2  Simple Kriging with Many Predictors

                   6.2.3  Interpretation of Prediction Weights

                   6.2.4  Construction of Prediction Intervals

                   6.2.5  Implementation of Simple Kriging Models

                   6.2.6  An Example of Simple Kriging

        6.3  The Ordinary Kriging Model

                   6.3.1  Best Linear Unbiased Estimation of the Mean

               6.3.2  Best Linear Unbiased Predictor of Y

                   6.3.3  Implementation of Ordinary Kriging

                   6.3.4  An Example of Ordinary Kriging

        6.4  Selection of Prediction Sets by Cross Validation

                   6.4.1  Log-Nickel Example

                   6.4.2  A Simulated Example

7. General Spatial Prediction Models


        7.1  The General Linear Regression Models

                   7.1.1  Generalized Least Squares Estimation

                   7.1.2  Best Linear Unbiasedness Property

                7.1.3  Regression Consequences of Spatially Dependent

                          Random Effects.

        7.2  The Universal Kriging Model

                   7.2.1  Best Linear Unbiased Prediction

                   7.2.2  Standard Error of Predictions

                   7.2.3  Implementation of Univesal Kriging

        7.3  Geostatistical Regression and Kriging

                   7.3.1  Iterative Estimation Procedure

                   7.3.2  Implementation of Geo-Regression

                   7.3.3  Implementation of Geo-Kriging

                 7.3.4  Cobalt Example of Geo-Regression

                 7.3.5  Venice Example of Geo-Regression and Geo-Kriging

APPENDIX TO PART II

         A2.1.   Covariograms for Sums of Independent Spatial Processes

          A2.3.  Expectation of the Sample Estimator under Sample Dependence

          A2.3.  A Bound on the Binning Bias of Empirical Variogram Estimators

          A2.4.   Some Basic Vector Geometry

          A2.5.  Differentiation of Functions

          A2.6.  Gradient Vectors

          A2.7.  Unconstrained Optimization of Smooth Functions

                     7.1  First-Order Conditions

                     7.2  Second-Order Conditions

                     7.3  Application to Ordinary Least Squares Estimation

          A2.8.  Constrained Optimization of Smooth Functions

                     8.1  Minimization with a Single Constraint

                     8.2  Minimization with Multiple Constraints

                     8.3  Solution for Universal Kriging

 

III. AREAL DATA ANALYSIS

 

1. Overview of Areal Data Analysis


        1.1  Extensive versus Intensive Data Representations
        1.2  Spatial Pattern Analysis

        1.3  Spatial Regression Analysis

2. Modeling the Spatial Structure of Areal Units


        2.1  Spatial Weights Matrices

               2.1.1  Point Representations of Areal Units
               2.1.2  Spatial Weights based on Centroid Distances

               2.1.3  Spatial Weights based on Boundaries

               2.1.4  Combined Distance-Boundary Weights

               2.1.5  Normalizations of Spatial Weights

        2.2  Construction of Spatial Weights Matrices

               2.2.1  Construction of Spatial Weights based on Centroid Distances
               2.2.2  Construction of Spatial Weights based Boundaries

3. The Spatial Autoregressive Model


        3.1  Relation to Time Series Analysis

        3.2  The Simultaneity Property of Spatial Dependencies

        3.3  A Spatial Interpretation of Autoregressive Residuals        

               3.3.1  Eigenvalues and Eigenvectors of Spatial Weights Matrices
               3.3.2  Convergence Conditions in Terms of Rho

               3.3.3  A Steady-State Interpretations of Spatial Autoregressive Residuals

4. Testing for Spatial Autocorrelation


        4.1  Three Test Statistics

               4.1.1  Rho Statistic

               4.1.2  Correlation Statistic

               4.1.3  Moran Statistic

               4.1.4  Comparison of Statistics

        4.2  Asymptotic Moran Tests of Spatial Autocorrelation

                 4.2.1  Asymptotic Moran Test for Regression Residuals

               4.2.2  Asymptotic Moran Test in ARCMAP

        4.3  Random Permutation Test of Spatial Autocorrelation

               4.3.1  SAC-Perm Test

               4.3.2  Application to English Mortality Data

5. Tests of Spatial Concentration


        5.1  A Probabilistic Interpretation of G*

        5.2  Global Tests of Spatial Concentration

        5.3  Local Tests of Spatial Concentration

               5.3.1  Random Permutation Test

               5.3.2  English Mortality Example

               5.3.3  Asymptotic G* Test in ARCMAP

               5.3.4  Advantage of G* over G for Analyzing Spatial Concentration

6. Spatial Regression Models for Areal Data Analysis


        6.1  The Spatial Errors Model (SEM)

        6.2  The Spatial Lag Model (SLM)

               6.2.1  Simultaneity Structure

               6.2.2  Interpretation of Beta Coefficients

        6.3  Other Spatial Regression Models

               6.3.1  The Combined Model

               6.3.2  The Durbin Model

               6.3.3  The Conditional Autoregressive (CAR) Model

 

APPENDIX TO PART III

         A3.1.  The Geometry of Linear Transformations

                   3.1.1 Nonsingular Transformations and Inverses

                   3.1.2 Orthonormal Transformations

          A3.2.  Singular Value Decomposition Theorem

                   3.2.1 Inverses and Pseudoinverses

                   3.2.2 Determinants and Volumes

                   3.2.3 Linear Transformations of Random Vectors

          A3.3.  Eigenvalues and Eigenvectors

          A3.4.  Spectral Decompostion Theorem

                   3.4.1 Eigenvalues and Eigenvectors of Symmetric Matrices

                   3.4.2 Some Consequences of SVD for Symmetric Matrices

                   3.4.3 Spectral Decomposition of Symmetric Positive Semidefinite Matrices

                   3.4.4 Spectral Decompositions with Distinct Eigenvalues

                   3.4.5 General Spectral Decomposition Theorem

   

IV. SOFTWARE 

     1.  ARCMAP

       1.1  Opening ARCMAP

       1.2  Tips for Using ARCMAP

              1.2.1    Importing Text Files to ARCMAP
              1.2.2    Changing Path Directories in Map Documents
              1.2.3    Making a Column of Row Numbers in an Attribute Table
              1.2.4    Masking in ARCMAP
              1.2.5    Making Spline Contours in Spatial Analyst
              1.2.6    Excluding Values from Map Displays
              1.2.7    Importing ARCMAP Images to the Web
              1.2.8    Adding Areas to Map Polygons
              1.2.9    Adding Centroids to Map Polygons

              1.2.10  Adding Coordinate Fields to Attributes of Point Shapefiles
              1.2.11  Converting Strings to Numbers in ARCMAP
              1.2.12  Displaying Proper Distance Units
              1.2.13  Editing Point Styles in ARCMAP
              1.2.14  Exporting Maps from ARCMAP to WORD
              1.2.15  Making Legends for Exported Maps
              1.2.16  Making Voronoi Tessellations in ARCMAP as Shapefiles
              1.2.17  Running Local G* Tests of Concentration in ARCMAP
              1.2.18  Joining Point Date to Polygon Shapefiles in ARCMAP
              1.2.19  Saving Map Documents with Relative Paths
              1.2.20  Increasing Unique Values for Editing Raster Outputs (in Version 9.3)

    2. JMPIN

       2.1  Opening JMP

       2.2  Tips for using JMP

              2.2.1   Printing Results from JMP
              2.2.2   Making a Random Reordering of Row Numbers 

    3. MATLAB

       3.1  Opening MATLAB

       3.2  Tips for using MATLAB

              3.2.1   Exporting Graphics from MATLAB to WORD
              3.2.2   Making Boundary-Share Weight Matrices in MATLAB
              3.2.3  
Making Boundary-Share Weight Matrices using ARCMAP and MATLAB
              3.2.4   Clipping Grids in ARCMAP for use in and MATLAB

              3.2.5   Exporting Data from MATLAB to ARCMAP

              3.2.6   Converting Boundary Shapefiles to MATLAB format.
                      

 REFERENCES

(return to contents)


Last modified: January 11, 2014