Spatial Socio-econometric Modeling (SSEM): A Low-Code Toolkit for Spatial Data Science and Interactive Visualizations Using R

Description

With the primary goal of expanding access to spatial data science tools, this book offers dozens of minimal or low-code functions and tutorials designed to ease the implementation of fully reproducible Spatial Socio-Econometric Modeling (SSEM) analyses. Designed as a University of Pennsylvania Ph.D. level course for sociologists, political scientists, urban planners, criminologists, and data scientists, this textbook equips social scientists with all concepts, explanations, and functions required to strengthen their data storytelling. It specifically provides social scientists with a comprehensive set of open-access minimal code tools to: •Identify and access place-based longitudinal and cross-sectional data sources and formats•Conduct advanced data management, including crosswalks, joining, and matching •Fully connect social network analyses with geospatial statistics•Formulate research questions designed to account for place-based factors in model specification and assess their relevance compared to individual- or unit-level indicators•Estimate distance measures across units that follow road network paths •Create sophisticated and interactive HTML data visualizations cross-sectionally or longitudinally, to strengthen research storytelling capabilities•Follow best practices for presenting spatial analyses, findings, and implications•Master theories on neighborhood effects, equality of opportunity, and geography of (dis)advantage that undergird SSEM applications and methods•Assess multicollinearity issues via machine learning that may affect coefficients' estimates and guide the identification of relevant predictors•Strategize how to address feedback loops by using SSEM as an identification framework that can be merged with standard quasi-experimental techniques like propensity score models, instrumental variables, and difference in differences•Expand the SSEM analyses to connections that emerge via social interactions, such as co-authorship and advice networks, or any form of relational data The applied nature of the book along with the cost-free, multi-operative R software makes the usability and applicability of this textbook worldwide. Preface Unique Contribution to Social Sciences SSEM Statistical Modeling Culture Level at Which This Book is Aimed References Acknowledgements Contents Acronyms List of Figures List of Tables Code Listings for Replication Exercises Part I Conceptual and Theoretical Underpinnings 1 SPlaces SPlaces: Spaces, Places, and Spatial Socioeconometric Modeling Spaces Places SPlaces Inequality in Mobility Prospects Measuring Inequality and Growing Inequality Neighborhood Effects and Concentration of (dis)Advantages Splace-Based Modeling Challenges Causality in Spatial Modeling Individual and Place-based Multicollinearity Closing Thoughts and Next Steps Next Steps Discussion Questions References 2 Operationalizing SPlaces Delimiting and Operationalizing Neighborhoods as Splaces Representing Physical Spaces and Nesting Structures Zooming in Across Administrative Boundaries Shapefiles as Spaces Elements of a Shapefile Place-Based Indicators Contributing to Building Splaces Neighborhood Operationalization and Disaggregation Data Point Differentiation Across Neighborhood Levels Illustration of Splaces and Data Point Gains Tradeoffs of Data Point Differences What Might be the Best Choice? Bringing Concepts, Shapefiles, and Place-Based Indicators Together ACS Published or Pre-tabulated Data Identifying Proxies for Poverty Identifying Proxies for Median Income Identifying Proxies for Unemployment Identifying Proxies for Housing Quality Identifying Proxies for Family Structure Closing Thoughts and Next Steps Next Steps Discussion Questions References 3 Data Formats, Coordinate Reference Systems, and Differential Privacy Frameworks Types of Geo-Referenced Data: Raster and Vector Data Raster Data Vector Data Point Geometries Line Geometries Polygon Geometries Geometries as Layers Vector to Raster Transformations and Vice Versa Vector to Raster Data Transformations Moving From Raster to Vector Data From Raster to Points From Raster to Polygons Coordinate Reference Systems Elements of CRS Implications of Distortions Resulting from Map Projections Why is CRS Harmonization Important? Commonly Used Coordinate Reference Systems Differential Privacy Framework (DPF) and Changes to Census Micro Data Are Differential Privacy and Synthetic Data the Same Privacy Protection Strategy? What are Differential Privacy Algorithms? Relevance of Differential Privacy For SSEM Strategies to Protect Privacy Methodological Implications of DPF for SSEM Next Steps Discussion Questions References Part II Data Science SSEM Identification Tools: Distances, Networks, and Neighbors 4 Access and Management of Spatial or Geocoded Data R Tutorial Installation R Infrastructure Code Rationale Reading Data from an External Source Creating Datasets from Within R Merging Joining Data Installing Packages Moving Forward Reading Polygon Shapefiles Reading Polygons at the Country Level Reading Polygons at the County Level Reading Polygons at the ZIP Code Tabulated Area (ZCTA) Level Reading Polygons at the Census Tract Level Reading Polygons at the Block Group Level Reading Line Shapefiles All Roads Shapefiles Primary Roads Shapefiles Primary and Secondary Roads Shapefiles Appending Polygon Shapefiles Reading Point Shapefiles Point Geocoding or Georeferencing Batch Geocoding Using Addresses in R From data.frame to sf Objects Point Batch Geocoding Using ZCTAs in R Crosswalking Lower to Higher Level Crosswalking Place-Based Data Access at the Polygon Level Applying for a Census API Key Poverty Proxy Median Income Unemployed in Labor Force (Civilian) Housing Quality: Plumbing Facilities Family Structure: Women Led Households Joining ACS Databases IRS Data Place-Based Data Access at the Point Level Joining Points with Polygons Data Closing Thoughts Next Steps Discussion Questions Replication Exercises References 5 Distances Distances Geolocated Data: Polygons, Points, or Both? Why is Distance Estimation Relevant for SSEM? Data Source and Data Requirements Projections, Distortions, and Bias Concerns? Network Analysis Tools and Data Transformations Approaches to Distance Connections Identification Multiple Sources of Points Matrix to Edgelist Transformations Transformations Using One Unit Type Transformations Using Two Unit Types Summary and Next Steps for Distance Calculations As the Crow Flies Distance Calculations From a Matrix to a List of Connections (Edgelist) with Distances ``As the Crow Flies'' Distances Including Multiple Unit Types Network Route Distance Calculations: As Humans Walk As Humans Walk Distances Between Two Points humanswalktwo Function Application humanswalktwo(...) Function Components Batch ``As Humans Walk'' Distances ``As Humans Walk'' Batch Data Requirements Network Transformations humanswalkbatch Function Application Batch Walking Distances Among Units of the Same Type Batch Walking Distances Among Units of Different Types Moving Beyond Single Counties and States Navigation/Travel Time Distances ``Travel Distance'' Data Format and Requirements Types of Travel Time Estimates Supported traveltimes Function Applications Comparing Travel Times with Google Maps Closing Thoughts Next Steps Discussion Questions Replication Exercises References 6 Geographical Networks as Identification Tools Neighboring Structures and Networks What is a Network and How is it Different from or Similar to Neighboring Structures? From Distances (or Travel Times) to Networks and Neighboring Structures Point-Based Network and Neighboring Structures Identification Rules Radius-Based Approach Kth Closest Neighbor(s) Approach Inverse Distances From Neighboring Structures to Weights Code Application Moving Forward and Beyond These Standard Identification Approaches Crow Flies Versus Road Networks Distances Crow Flies Applying Road Networks Distances: ``As Humans Walk'' Using our Own Network Distances (and/or Travel Times) to Identify Neighboring Structures rad Function klosest Function radkthrow Function radkthinv Function Moving Forward Identifying Neighboring Structures Among Different Types of Units Identifying the Local Presence of Units of Different Type What is Commuting Distance? Visualization Indirect Neighboring Structures Transformation Application to Real Data Moving Forward and Feedback Loops Two-Mode Kth Closest Identification and Selection Identification Application Adding a Threshold to Nearby Limits and Place Heterogeneities Feedback Loops and Self-selection Polygons and Matrices of Influence Rook's Bishop's Queen's Application Higher Order Neighbors Closing Thoughts Next Steps Discussion Questions Replication Exercises References Part III SSEM Hypothesis Testing of Cross-Sectional and Spatio-Temporal Data and Interactive Visualizations 7 SODA: Spatial Outcome Dependence or Autocorrelation SODA: Spatial Outcome Dependence or Autocorrelation Why Is SODA Statistically Concerning? Assessing SODA Based on Polygon Data Moran's I Regression Approach Moran's I Code Application with Polygon Data Is the First Order Neighboring Structure Enough? Combining Higher Order Neighboring Structures Decision Selection Process to Model Higher Order Neighbors Assessing SODA Based on Point Data Machine Learning Tools to Assess SODA Decadence Moran's I Code Application with One-Mode Point Data Data Source and Outcome of Interest Neighboring Structures and Weight Matrix Analytic Steps Code Application Moran's I Code Application with Two-Mode Point Data Two- To One-Mode Transformations and Rationale Analytic Steps Causal Chains Through Spillovers in SSEM Local Moran's I Visualizing Local Moran's I Quadrant Representation Map Representation Code Application Local Moran's I Polygon Data Code Application Local Moran's I One-Mode Point Data Code Application Local Moran's I Two-Mode Point Data To Retain or Exclude Neighborless Units Code Application to Exclude Neighborless Units Social Outcome Dependence or Autocorrelation: SODA 2.0 Relationships in SODA 2.0 Application of SODA 2.0 Adjacency List to Edgelist Transformation Author's Individual Publication Record SODA 2.0 Interpretation Moving Forward with SODA 2.0 Next Steps Discussion Questions Replication Exercises References 8 SSEM Regression Based Analyses Residual SODA and the Importance of Spatial Regression Modeling SODA Mechanisms in Regression Residuals Testing for RSODA Simultaneous Autoregressive (SAR) Modeling Mechanisms and Implications of RSODA SAR Application to Polygon Data Assessing Whether RSODA was Handled Building a SAR Model While Addressing Place-based Multicollinearity Application of Feature Selection Via Random Forests Application Simultaneous Autoregressive Models Revisiting the Notion of Splaces and Data Point Gains SAR Application to Two-Mode Point Data Data Preparation and Transformations Outcome Indicators and Feature Selection Rationale Two-mode to One-mode Transformations Neighborless Units and Decision Making Feature Selection with Point Data Building SAR Model While Addressing Place-based Multicollinearity Multilevel SAR Models Multilevel Data How Does SAR differ from Multilevel SAR? Statistical Description of Multilevel SAR Polygon Matrices of Influence 數瑥浤慳栠Higher Level Matrix M Point Matrices of Influence 數瑥浤慳栠Lower-Level Matrix W Δ Matrix to Account for Fixed Group (or Nesting) Effects Bringing the the Three Matrices Together Multilevel SAR Function Application Two-Mode Multilevel SAR Multilevel SAR Results: Two Mode Multilevel SAR Results: One Mode SAR or Multilevel SAR Testing for Spatial Heterogeneity Via Geographically Weighted Regression How Does SAR differ from GW Approaches? Distance and Travel Time Matrices and Kernel Functions Do we Need GW?: GW Multiscale Summary Statistics WG Summary Execution Geographically Weighted Regression and Visualization Bootstrap Strategy WG Regression Execution Mapping of GW Results Spatio-Temporal SAR: A Difference in Differences Application Spatio-Temporal Data or Panel Data with Spatial Information Testing for RSODA in Panel SAR SAR Panel Set Up SAR Panel Data Source and Setting Identification Falsification Test Identification SAR Panel Application SAR Panel Function Falsification Tests SAR and Multilevel SAR with Social Data Multilevel SAR Constrains for SODA 2.0 Social Multilevel SAR socialmultilevelSAR(...) Application Closing Thoughts and Next Steps Discussion Questions Replication Exercises References 9 Visualization, Mining, and Density Analyses of Spatial and Spatio-Temporal Data SSEM Visualizations Polygon Data Visualization polymap(...) Function Application Exploratory Spatio-Temporal Data Mining and Visualization spatiopanelvisual(...) Implementation Point Data Visualization pointmap(...) Function Application Geospatial Point Density Methodological Approach What Questions May We Address with Geospatial Point Density? Findings Research Question 1 Findings Research Question 2 Code Application for the Maps Next Steps in Gesopatial Point Density Geographical Network Visualizations Data Sources Preparation Rationale geographicalnetworks(...) Application Two-Mode Networks Application Two- to One-Mode Transformation One-Mode Geographical Networks Procedures Closing Thoughts Discussion Questions Replication Exercises References 10 Final Words References Glossary Index