1.2 A Quick Tour of GeoDa

Before delving into the specifics of particular methods, I provide a broad overview of the functionality and overall organization of the GeoDa software. The complete toolbar with icons corresponding to a collection of related operations is shown in Figure 1.1. Each icon is matched by a menu item, detailed in Appendix B. The menu and user interface can be customized to several languages (details are in Appendix A). The default is English, but options are available for Simplified Chinese, Russian, Spanish, Portuguese and French, with more to come in the future.

With each toolbar icon typically corresponds a drop-down list of specific functions. The structure of the drop-down list matches the menu sub-items (Appendix B).

Figure 1.1: GeoDa Toolbar Icons

The organization of the toolbar (and menu) follows the same logic as the layout of the parts and chapters in the two books. It represents a progression in the exploration, from left to right, from support functions to queries, description and visualization, and more and more formal methods, ending up with the estimation of actual spatial models in the regression module (not covered here).

A brief overview of each of the major parts is given next. This also includes the spatial clustering functionality, which is discussed more specifically in Volume 2.

1.2.1 Data Entry

The three left-most icons, highlighted in Figure 1.2, deal with data entry and general input-output. This includes the loading of spatial and non-spatial (e.g., tabular) data layers from a range of GIS and other file formats (supported through the open source GDAL library). In addition, it offers connections to spatial data bases, such as PostGIS and Oracle Spatial. It also supports a Save As function, which allows the software to work as a GIS file format converter. Further details are provided in Chapter 2.

Figure 1.2: Data Entry

1.2.2 Data Manipulation

Functionality for data manipulation and transformation is provided by the Table icon, highlighted in Figure 1.3. This allows new variables to be created, observations selected, queries formulated and includes other data table operations, such as merger and aggregation, detailed in Chapter 2.

Figure 1.3: Data Manipulation/Table

1.2.3 GIS Operations

Spatial data operations are invoked through the Tools icon, highlighted in Figure 1.4. These include many GIS-like operations that were added over the years to provide access to spatial data for users who are not familiar with a GIS. For example, point layers can be easily created from tabular data with X,Y coordinates, point in polygon operations support a spatial join, an indicator variable can be used to implement a dissolve application, and reprojection can be readily implemented by means of a Save As operation. Specific illustrations are included in Chapter 3.

Figure 1.4: GIS Operations/Tools

1.2.4 Weights Manager

The Weights Manager icon, Figure 1.5, contains a final set of functions that are in support of the analytical capabilities. It gives access to a wide range of weight creation and manipulation operations, discussed at length in the chapters of Part III. This includes constructing spatial weights from spatial layers, as well as loading them from external files, summarizing and visualizing their properties, and operations like union and intersection.

Figure 1.5: Weights Manager

1.2.5 Mapping and Geovisualization

The mapping and geovisualization functionality is represented by four icons, highlighted in Figure 1.6: the Map icon, Cartogram, Map Movie and Category Editor. The mapping function supports all the customary types of choropleth maps, as well as some specialized features, such as extreme value maps, co-location maps and smoothed maps for rates. The cartogram is a specialized type of map that replaces the actual outline of spatial units by a circle, whose area is proportional to a given variable of interest. Animation, in the sense of moving through the locations of observations in increasing or decreasing order of the value for a given variable is implemented by means of the map movie icon. Finally, the category editor provides a way to design custom classifications for use in maps as well as in statistical graphs, such as a histogram. Details are provided in Chapters 4 through 6.

Figure 1.6: Mapping and Geovisualization

1.2.6 Exploratory Data Analysis (EDA)

The next eight icons, grouped in Figure 1.7, contain the functionality for exploratory data analysis and statistical graphs. This includes a Histogram, Box Plot, Scatter Plot, Scatter Plot Matrix, Bubble Chart, 3D Scatter Plot, Parallel Coordinate Plot and Conditional Plots. These provide an array of methods for univariate, bivariate and multivariate exploration. All the graphs are connected to any other open window (graph or map) for instantaneous linking and brushing. This is covered in more detail in Chapters 7 and 8.

Figure 1.7: Exploratory Data Analysis

1.2.7 Space-Time Analysis

The exploration of space-time data, treated in Chapter 9, is invoked by means of the icons on the right, highlighted in Figure 1.8. This includes a Time Editor, required to transform the cross-sectional observations into a proper (time) sequence. In addition, the Averages Chart implements a simple form of treatment analysis, with treatment and controls defined over time and/or across space.

Figure 1.8: Space-Time Analysis

1.2.8 Spatial Autocorrelation Analysis

Spatial autocorrelation analysis is invoked through the three icons highlighted in Figure 1.9. The first two pertain to global spatial autocorrelation. The left-most icon corresponds to various implementations of the Moran scatter plot (Chapters 13 and 14). The middle icon invokes nonparametric approaches to visualize global spatial autocorrelation, as a spatial correlogram and distance scatter plot (Chapter 15).

The third icon contains a long list of various implementations of local spatial autocorrelation statistics, including various forms of the Local Moran’s I, the Local Geary c, the Getis-Ord statistics, and extensions to multivariate settings and discrete variables. The local neighbor match test is a new method, based on an explicit assessment of the overlap between locational and attribute similarity. Details are provided in the chapters of Part V.

Figure 1.9: Spatial Autocorrelation Analysis

1.2.9 Cluster Analysis

Finally, cluster analysis is invoked through the icon highlighted in Figure 1.10. An extensive drop down list also includes the density based cluster methods DBScan and HDBScan, which are treated in this volume under local spatial autocorrelation (Chapter 20).

The other methods are covered in Volume 2. They include dimension reduction, classic clustering methods and spatially constrained clustering methods. The last items in the drop-down list associated with the cluster icon pertain to the quantitative and visual assessment of cluster validity, including a new cluster match map (see Volume 2).

Figure 1.10: Cluster Analysis