1 Data Management and Basic Mapping with GeoDa Reinhold Kosfeld Institute of Economics University of Kassel August 1, 2011 Contents 1.Introduction 2.GeoDa Main Menu and Toolbar 3. Opening a Project 4. Merging a Shape and Data File Excursus ● Creating a Shape File ● Creating a dBase Data File 5. Creating Basic Choropleth Maps 6. Basic Table Operations and Linking 7. Creating a Contiguity Weights File 8. Weight Characteristics and Linking 2 1. Introduction GeoDa is a collection software tools for spatial data analysis developed under the lead of Luc Anselin (University of Illinois and Center of Spatially Integrated Social Sciences) at the Center of Spatially Integrated Social Sciences (CSISS). Geographical data must relate to discrete spatial units like regions, districts, counties,or zip codes. This kind of data is usually termed lattice data. Actually, GeoDa does not allow the analysis of geostatistical data (continuous surface data) or point pattern data (locations of events). GeoDaprovides a user-friendly and graphical interface to a variety of methods of standard exploratory and spatial data analysis. For standard exploratory data analysis (EDA), graphical devices like histograms, box plots and diverse types of scatter plots are available. Descriptive spatial dataanalysis provides for example measures of spatial autocorrelation and indicators of spatial outliers. Local measures will usually be mapped for the whole study region.Advanced technical capabilities like linking and brushing are implemented in GeoDa. Inferential spatial analysis is in particular involved with spatial autocorrelation and regression analysis. For different purposes it will be of advantage to map local pvalues or regression residuals can also be mapped for the system of regions.In the present version of GeoDA,OpenGeoDa 0.9.9.12, regression analysis can be accomplished for a cross-section of regions. Up to now, GeoDa is not geared to analysing spatial panels. GeoDa can be downloaded at no charge from the GeoDa Center website http://geodacenter.asu.edu/software/downloads. All the user has to do is to register. Once a user account is created, the software can be downloaded. Tutorials and other materials related to GeoDa are available on http://geodacenter.asu.edu/learning/ tutorials. A comprehensive survey on GeoDa is covered by the webbook “Exploring Spatial Data with GeoDa: A Workbook”byAnselin. This tutorial deals with different types of files necessary for mapping and spatial data analysis with GeoDa: - input data file, polygon shape file, weight file. The management of these files in GeoDa is illustrated by examples from applied research of the author in the field of regional economics. 3 2.GeoDa Main Menu and Toolbar Start GeoDa by double-clicking on the GeoDa icon on the desktop or choosing GeoDa from the list of programs displayed after clicking on the start icon. The Main Menu provides utilities for data management and techniques for mapping and non-spatial and spatial statistical analysis under the menu items Map, Explore, Space and Method. Most of the functions can also be invoked by the toolbar buttons. Figure 2.1: GeoDa main menu and toolbar From the Map menu, for example, quantile and box maps are available. Likewise different types of rate maps as special cases of choropleth maps can be constructed (Figure 2.2).Note that all items greyed out which means that the methods are not available without further ado. Figure 2.2: Map menu These mapping facilities can also be invoked by clicking on the Maps and Smoothing button in the toolbar (Figure 2.3). Without defining a project, no list of methods is visible. 4 Figure 2.3: Maps and Smoothing toolbar button The Explore menu provides methods of exploratory data analysis (EDA) like the histogram, scatter plot and box plot (Figure 2.4).Toolbar buttons are available for each of the EDA techniques. Figure 2.4: Explore menu Spatial autocorrelation analysis can be invoked by the Space Menu (Figure 2.5). The global and local Moran coefficient is as well available as local Getis-Ord statistics (G statistics). All spatial autocorrelation coefficients and testscan be also called up by toolbar buttons. Figure 2.5: Space menu The Methods Menu involves standard least-squares and spatial regression models. Regression analysis with GeoDa also involves several tests on coefficients and residuals. In particular tests on spatial autocorrelation of the disturbances are accomplished. Regression analysis cannot be invoked from the toolbar. 5 Figure 2.6: Methods menu While the Table menu provides facilities for managing tables and performing table calculations, the Tools menu especially contains capabilities for creating spatial weights, creating and converting shape files and exporting data. The items of Options menu depend on the chosen technique of data analysis. Currently the Help menu only gives information on the GeoDa Center and the current version of GeoDa. All items greyed out in the submenus require a project to be defined in the first place. A project consists of a shape file covering geographical information on a system of regions that is usually extended by data from a special field of application.In the next section we will consider, how a project is opened from the File Menu or the Open Project toolbar button. 3. Opening a Shape File You ordinarily start data analysis with GeoDa by opening a project from the File menu(Figure 3.1) or clicking on the Open Project button of the toolbar (Figure 3.2). Figure 3.1: Open Shape File from the File menu Figure 3.2: Open Project from toolbar 6 In either case the window Choose a Shape file to open window appears (Figure 3.3). Select the shape file DKR439.shp that contains geographical information of all 439 German districts. Figure 3.3: Choose a Shape File to open window . . After clicking on the Open button, a blank map will appear showing all 439 districts of Germany on the right hand side. On the left hand side you find the legend pane. Figure 3.4: Blank map of German districts1 Both panes can be resized by dragging the separator between them. The settings can be changed by right clicking in the map window. In the right pane a different colour can be selected from a palette of colours e.g. for the map, the background or highlighting regions. There is also the possibility to change the shape of the selection 1 © EuroGeographics for the administrative boundaries. 7 tool. By right clicking in the left pane another background colour for the legend can be selected. Exercises 3.1 Resize the legend pane so that the complete file name is readable. 3.2 Display the map of the German districts in red colour. 3.3 Choose the colour yellow for the background of the map pane. 3.4 Choose grey as the colour for the background of the legend pane. 3.5 Portray only the isle of Rugia in the map pane. 4. Merging a Shape and Data File When a shape file is loaded, the complete main menu and toolbar is active. At the start the shape usually only contains geographical information with an identifier for the regions. For spatial data analysis, this information must be combined with datafrom the field of application. The data must be available in a dBase file (.dbf file). We will merge employment data set Besch06EKr with the shape file DKR439.shp of the 439 German districts. This data set contains sectoral employment figures at the district level for research and development (R&D) intensive industries and the entire manufacturing industry. Additionally. Population figures are included: DG24: Chemical Industry DK29: Manufacture of machinery and equipment DL: Manufacture of electrical and optical instruments DL30: Manufacture of office machinery and computers DL31:Manufacture of electrical machinery and apparatus DL32:Manufacture of radio, television and communication equipments and apparatus DL33:Manufacture of medical, precision and optical instruments, watches and clocks DM34: Manufacture of motor vehicles, trailers and semi-trailers D_VG: Manufacturing Industry EINW: Population Load the shape file DKR439.shp. Open the Table Menu and click on Merge Table Data (Figure 4.1). 8 Figure 4.1: Merge Table Data from the Table menu A table with the geographical variables for the 439 German district (Figure 4.2) appears along with the Merge Table Data window (Figure 4.3). The table associated with the shape file consists of geographical variables: - SP_ID (district identifier), - NUTS_ID (NUTS identifier), - SHAPE_LENG (average district length) - SHAPE_AREA (district area) - KREISCODE (official district code (identical with SP_ID) - NAME (district name) - ZUORDNUNG (rural or urban district, federal city state). The variables AREA an LEN are set equal to zero. The values of the area and length of the districts are listed in SHAPE_AREA and SHAPE_LENG. Figure 4.2: Table with Geographical Variables from the Shape File DKR439.shp 9 Navigate in the familiar window structure of the Merge Table Data window until the input file Besch06EKr.dbf is located (Figure 4.3). The Key is an identifier variable that must be present in both the shape and data file to uniquely match the spatial units. Leave the Key to its default KREISCODE. Figure 4.3: Merge Table Data window Select the variables of the variables to be included in the shape file. First, click on the upper double arrow button (>>)to include all variables. Next mark the variable KREISCODE and click on the upper single arrow button (<). Figure 4.4: Select variables in the Merge Table Data window 10 The Key variable need not twice be included in the new shape file. Now click on the Merge button to join both files. Figure 4.5 displays the joint table with combined geographical information and sectoral employment data for the 439 German districts. Figure 4.5: Joint table after Merging Table Data of the shape file DKR439.shp and data file Besch06EKr.dbf Before storing the extended shape file permanently, the redundant variables AREA and LEN should be deleted. For this, invoke the Table function from the main menu and select the item Delete Column (Figure 4.6). Figure 4.6: Delete Column from Table menu The Remove Column window appears (Figure 4.7). Choose the variable AREA from the pull-down menu in the field Pick the column name and click on the Delete button: 11 Figure 4.7: Remove Column window The variable AREA is removed from the table. Delete the variable LEN in the same way. Now open the Table Menu again to store the extended shape file permanently. Choose Save as New Shape File … from the Table menu (Figure 4.8): Figure 4.8: Save as New Shape File … from the Table menu The Output Shp file window appears in the familiar Windows dialog. Change the displayed file name to DKR439Besch06E.shp. Finally, store the new shape file by clicking the Store button: Figure 4.9: Storing the extended Shape file DKR439Besch06E.shp 12 The new shape file DKR439Besch06E.shp with sectoral employment data is now available for data analyses and mapping with GeoDa. Excursus ● Creating a Shape File Technically, a map of a system of adjacent regions can be viewed as an irregular lattice. In order to map the boundaries of the regions in GeoDa, a polygon shape file is required. The shape file will be created by GeoDa from a text file that contains polygon boundary coordinates. Figure E4.1 shows the structure of the input text file DKR439.txt from which the shape file DKR439.shp for the German districts couldprincipally be created.2 Figure E4.1: Text file with coordinates of the boundaries of the German districts The header line contains the number of districts followed by polygon identifier (= region identifier). Both figures are separated by a comma. On the following lines, the 2 Here we used a completed shape file of German districts from EUROSTAT. 13 polygons (= regions) are geographically defined. The first line for each polygon contains consists of the identifier and the number of points, separated by a comma. The latter are defined by coordinate pairs that are listed in the following lines.In the case of the German districts we make use of the latitude and longitude coordinates3 ● Creating a dBase File In GeoDa, data is read from a dBase file format (DBF file format). Since data is often stored in spreadsheet files (.xlsx files), these files have to be converted in a DBF format. This could be done by special conversion programs. Alternatively, one can use the open source software R for mathematical and statistical computing and programming which is free available on the homepage of the R project http://www.rproject.org. Assume that the data is available in the Excel spreadsheet file Besch06EKr.xlsx. Open the file Besch06EKr.xlsx (Figure E4.2). Figure E4.2: Employment data in the Excel file Besch06EKr.xlsx Store it in Excel as a CSV file (text file with comma separated values) Besch06EKr.csv. 3 For illustrative examples, polygons can be defined by arbitrary x and y coordinates. 14 Figure 4.3: Windows dialog for saving data in CSV format in Excel At last run the following commands in R to create the DBF file Besch06Kr.dbf: >library(foreign) > Besch06EKr = read.csv2(file="Besch06EKr.csv") >write.dbf(Besch06EKr,file="Besch06EKr.dbf") Exercises Five regions are arranged in the following form: 1 2 4 3 5 The x and y coordinates of the regions are given as follows: Region 1: 3.0,6.0; 8.0,6.0; 8.0,4.0; 5.0,4.0; 3.0,4.0 Region 2:8.0,6.0; 11.0,6.0; 11.0,5.0; 11.0,4.0; 8.0,4.0 Region 3:5.0,4.0; 8.0,4.0; 11.0,4.0; 11.0,3.0; 11.0,2.0; 5.0,2.0 Region 4:11.0,5.0; 15.0,5.0; 15.0,3.0; 11.0,3.0; 11.0,4.0 Region 5:15.0,5.0; 17.0,5.0; 17.0,3.0; 15.0,3.0 15 Regional data is available on the unemployment rate (u) and the vacancy rate (v): Region 1 2 3 4 5 u 6 8 8 11 12 v 3 3 2 1 1 4.1 Create a shape file region5_ppt.shp for the five adjacent regions! 4.2 Create a dBase file region5_uvdata.dbf with a matching identifier! 4.3 Merge the shape file region5_ppt with the dBase file region5_uvdata and store the extended shape file under the name region5uv.shp! 5.CreatingBasic Choropleth Maps Open the shape file DKR439Besch06E.shp by selecting File > Open Shape File … from the GeoDa menu. In this exercise, we wish to examine the spatial distribution of employees in R&D industries and the manufacturing industry as a whole with the aid of the GeoDa mapping possibilities. Figure 5.1 display different types of choropleth maps available under the Map menu in GeoDa. Choose the item Quantile as the simplest type of a choropleth map. Figure 5.1: Selecting the Quantile map from the main menu 16 The Variable Settings dialog requests the choice of a variable for mapping. Select D_VG in the field 1st Variable (X) for mapping the employees in the manufacturing sector (Figure 5.2). The field 2nd Variable (Y) is not needed for Quantile map. You could check the box Set the variables a default, if you intend to use the same variable in repeated analyses. Leave the box unchecked and click OK. Figure 5.2: Variable Settings for Quantile map At last you are asked for the number of classes (Figure 5.3). With the default number of 4 classes a quartile map is created. Figure 5.3: Number of classes for Quantile map After clicking OK, a quartile map for the employees in the manufacturing industry is created. Drag the separator in the map window to the right to make the full legend visible. Figure 5.3 shows the quartile map of the variable D_VG with a legend. The legend first names the variable D_VG followed by the colours used in the map for the classes. The classes are represented by intervals followed by frequencies in parentheses. Each class involves approximately 110 observations. 17 Figure 5.3: Quartile map of employees in manufacturing industry In order to focus the attention on outliers in the spatial distribution of a georeferenced variable, GeoDa provides specialized maps like the Percentile and Box map. At first take a look at the Percentile map that sorts the regions in classes of relative frequencies of < 1%, 1 – 10%, 10 – 50%, 50 – 90%, 90 – 99% and > 99%. A Percentile map is invoked by selecting Map > Percentile in the GeoDa menu or clicking on the Maps and Smoothing button and choosing Percentilefrom the toolbar. The same Variable Settings dialog as for the Quantile map appears. After selecting the variable D_VG and clicking OK, GeoDa displays the Figure 5.4: Percentile map of employees in manufacturing industry 18 corresponding Percentile map. Drag the separator between the two panes so that the legend is fully in view. Potential outliers at the lower tail of the distribution are coloured in dark blue and those at the upper tail in dark red. Figure 5.4 points to four (five) potential outlying regions with low (high) employment figures. Another outlier map is the Box map. This is essentially a quartile map extended by an outlier identification rule. Outliers are regions with values that are more than 1.5 or 3.0 times the quartile range apart from the 1st or 3rd quartile of the mapped variable. Select Box Map from the Map menu with a hinge of 1.5 (Figure 5.4) or click on the Maps and Smoothing button and choose Hinge = 1.5. Figure 5.4: Selecting the Box Map from the main menu Figure 5.5: Selecting the Box Map form the toolbar This opens a familiar Variable Settings dialog familiar from the Quantile and Percentile map. Select again the variable D_VG and click OK.The resulting Box map is shown in Figure 5.6. 19 Figure 5.6: Box Map of employees in manufacturing industry As in the Percentile map, lower (upper) outliers are coloured in dark blue (dark red). Note that none of the potential outlying regions in the lower tail of the distribution of the employees in the manufacturing industry identified by the Percentile map are confirmed as outliers according to the identification rule (Hinge = 1.5) use in the Box map. On the other hand, it seems to be strange that the Box map identifies 33 upper outliers. Such problems often come along with so-called extensive variables that are not related to the “sizes” (area, population, etc.) of the regions. For choropleth mapping the use of intensive variables like rates and densities is usually more appropriate. In the next section we show how such kind of variables can be calculated in GeoDa. Exercises 5.1 Create a quartile and a quintile map for the number of persons employed in the electrical and optical industry (DL)! Interpret the maps! 5.2 Create a Percentile map for the variable DL! Which problem occurs when constructing this map? 5.3 Create Box maps for hinges of 1.5 and 3.0 for the variable DL! Compare the Box maps with each other! 5.4 In what way does a Percentile map differ from a Box map? Explain the absence of lower outliers with both the Percentile and Box map! 20 6.Basic Table Operations and Linking In GeoDa, the project with the shape file DKR439Besch06E.shp is already opened. Create once again the Percentile map Map>Percentile for the variable D_VG. Click on the Open Table button in the toolbar (Figure 6.1). Figure 6.1: Open Table button of the toolbar In Figure 6.2, an excerpts of the table associated with the project DKR439Besch06E is shown. Figure 6.2: Table from dBase file DKR439Besch06E.dbf … In order to identify the four potential upper outliers in the Percentile map, press SHIFT-CTRL and select the regions coloured in dark red by clicking on them. The selected regions are marked by a cross-hatch pattern (Figure 6.3). By right-clicking on the map or invoking theOptions menu you can change the colour of the crosshatch: Options > Color > Highlight Color. 21 Figure 6.3: Cross-hatched upper outlying regions in the Percentile map The selected districts are highlighted gray in the table (Figure 6.4). This feature is called “linking”. It does not only pertain to tables but also to all map and other graphs in GeoDa. Figure 6.4: Selected districts in the linked table . . . . . . . . . . . . . . . . . . 22 From Figure 6.4, the district names of the potential outlying regions are visible: Wolfsburg (3103)with 54733 D_VG employees, Märkischer Kreis (5962) with 64961 D_VG employees, Berlin (11000) with 103866 D_VG employees, Hamburg (2000)with 66334 D_VG employees, und Stuttgart (8111)with 71777 D_VG employees. Right click on the table or select the Table menu to move the upper potential outliers to the top: Table > Move Selected to Top Figure 6.5: Selected districts at the top of the linked table Clear the selection by clicking anywhere outside the map or choosing the Table menu: Table > Clear Selection You restore the original order of the observations in the table by double clicking on the header of left-most column with the sequence numbers. We have already seen how the table can be changed by deleting columns i.e. variables. The GeoDa table functionality also allows for adding new columns and transforming current variables. In regional economics, concentration of industries and specialisation of regions are often measured in relation to overall manufacturing activity. With this intention, we wish to analyse the spatial distribution of the sharenumber of persons employed by industry. Select the item Add Column in the Table menu (Figure 6.6) or by right-clicking on the table. 23 Figure 6.6: Add Column from Table menu This opens an Add Column dialog box with three fields (Figure 6.7). Enter DL_VG for the new column (new variable) in the Name field. Accept the default float in the Type field and choose EINW in the After field. Figure 6.7: Add Column dialog box After clicking on Add a new column with the header ANTDL is added right next to the EINW column. Next select Field Calculation from the Table menu (Figure 6.8) or by right-clicking on the table. Figure 6.8: Select Field Calculation from the Table menu 24 At the start the Filed Calculation with a dialog for Unary Operations i.e. operations with only one operand appears. Note the change of the dialog when clicking on the Binary Operations tab. Select ANTDL as the Result variable from the scroll-down menu on the left-hand side of the equation. Specify the division of Variable 1 DL by Variable 2 D_VG on the right-hand side as shown in Figure 6.9. Figure 6.9: Field Calculation window By clicking on Apply the zero entries in the ANTDL column are replaced shares employed persons between zero and one (Figure 6.10). Figure 6.10: Calculated shares of employed persons in column ANTDL 25 …. Using the new intensive variable ANTDL we wish to identify potential upper outliers with regard to manufacture of electrical and optical instrument. Select Map > Percentile from the main menu and choose ANTDL in the Variable Settings dialog as the 1st Variable. Ignore the warning regarding the uniquely sorting of records break points, as this only pertains to the lower tail of the distribution. Click on the 5 red coloured areas in the Percentile map. Figure 6.11 exhibits the location the five selected districts with a cross-hatch pattern. Figure 6.11: Potential upper outliers in Percentile map Click on the Open Table button in the toolbar and select Table > Move Selected to Top 26 to identify the districts with a high share of employees manufacture of electrical and optical instruments. According to linking, these districts are highlighted in grey in Figure 6.12. Figure 6.12: Potential upper ANTDL outliers in linked table … In order to verify whether these districts are outliers according to the HINGE = 3.0 identification rule, select Map > Box Map > Hinge = 3.0 in the GeoDa main menu. Choose ANTDL in the Variable Settings dialog and click OK. Mark the red coloured areas. The four upper outlying regions are evidenced by a cross-hatch in Figure 6.12. Figure 6.12: Outlying districts in Box Map with Hinge = 3.0 27 In Figure 6.13 the four outlying regions are place at the top of the table. Figure 6.13: Upper ANTDL outliers (Hinge = 3.0) in linked table … The identification rule (Hinge = 3.0) confirmed the districts Greifswald, Freiburg, Munich and Jena as upper outlying regions with regard to manufacture of electrical and optical instruments (DL industry). Their employment shares for this industry are more than 3 times quartile range away from the 3rd quartile. By contrast, the potential outlier Karlsruhe does not prove to be an outlier according to the applied identification rule with respect to the DL employment share. Exercises 28 6.1 Add two new columns ANTED_VG and ANTFD_VG to the table and calculate the rate and density of total employment in the manufacturing industry! 6.2 Create Percentile maps for the variables ANTED_VG and ANTFD_VGand locate the upper potential outliers! 6.3 Identify the selected districts in the linked tables of the Percentile maps created in exercise 6.2! 6.4 Create Box Maps with Hinge = 3.0 for the variables ANTED_VG and ANTFD_VG and locate the outlying areas! 6.5 Identify the selected districts in the linked tables of the Box Maps created in exercise 6.4! 7. Creating a Contiguity Weight File In GeoDa spatial weights can be constructed for a shape file that contains information on the contiguity structure and distances between regional centres. This information allows for creating both contiguity- and distance-based spatial weights. However, distances can actually only be used for defining regional neighbourhoods. Open the data shape file DKR439.shp with the variable KREISCODE as the key. Select Tools > Weights > Create from the main menu (Figure 7.1)or click on the Weights Create button in the toolbar (Figure 7.2). Figure 7.1: Weights creation from Tools menu Figure 7.2: Weights creation from toolbar 29 A Weights File Creating dialogwith several options appears. Specify DKR439 as the Input Shape File and choose KREISCODE as the Weights File ID Variable from the underlying scroll-down menu in the upper part of the window. Next, the type of spatial weights has to be specified. As we opt for contiguity-based weights, the field Distance Weight need not be considered. The field Contiguity Weight offers the choice between Queen or Rook Contiguity as well as different orders of contiguity. Check the radio button to the left of Rook Contiguity to define the neighbourhood between regions by common borders. Accept the default value of one for the 1st order of contiguity and click create. Figure 7.3: Weights File Creation This brings upthe Choose an output weights file name dialog(Figure 7.4). Figure 7.4: Choose an output weights file name 30 SelectDKR439 as the name for the spatial weights file. The GAL format of the weights file allows editingit with any text editorlike a simple text file (.txt). After storing the GAL file, the message ‘Weights file “DKR439.gal’ successfully created” appears. Click OK and Close the Weights File Creation window. The structure of the GAL file appears in Figure 7.5. The header line contains a placeholder (0), the number of regions (439), the name of the shape file (DKR439) Figure 7.4: Structure of GAL weights file and the name of the ID variable (KREISCODE). The other lines include regionspecific information. For each region two lines are reserved. While the first one 31 contains the region codes and the number of neighbours, the second one gives the region codes for the neighbours. Exercises 7.1 Create contiguity-based weights of order 1 for the shape file region5_ppt.shp on the basis of the rook criterion! 7.2 Edit the created GAL file region5_ppt.gal and interpret the contents! 7.3 Create contiguity-based weights of order 2 for the shape file region5_ppt.shp on the basis of the rook criterion! Interpret the contents of the created GAL file region5_ppt2.gal! 8. Weight Characteristics and Linking Before using the weights in spatial data analysis, they should be checked for their characteristics. Special attention should be drawn to undesirable features such as unconnected regions (“islands”). This can be done by examining the frequency distribution of the number of neighbours. In GeoDa, this distribution is displayed in form of a histogram that is generated by selecting Tools > Weights> Properties in the main menu (Figure 8.1) or clicking on the Weights neighbor histogrambutton in the toolbar (Figure 8.2). Figure 8.1: Weights Properties from the Tools menu Figure 8.2: Weights neighbors histogram from toolbar 32 This opens the WEIGHT CHARACTERISTICS box where the weight file has to be specified. Navigate in the working directory for the weight file DKR439.gal and click OK (Figure 8.3). Figure 8.3: Open weight file in the WEIGHT CHARACTERISTIC box The histogram of the number of neighbours is depicted in Figure 8.5. The height of the bars reflect the numbers of regions with 0, 1, …, 12 neighbours. The range of the number of neighboursis shown together with the absolute frequencies (in parentheses) in the legend. One region is unconnected i.e. is an “island”, whereas one other region has the maximum number of 12 neighbours. Figure 8.4: Histogram of number of neighbours with one “island” 33 The unconnected region is the isle of Rugia with the region code (KREISCODE) 13061. Although the isle of Rugia is an unconnectedregion in the map, it is actually connected by a bridge with the city of Stralsund (region code 13005). This information has to be included in the weights file (.gal) manually. The isle of Rugia has one and the city of Stralsund two neighbours. Update the weights file manually as in Figure 8.5 and store the changes in the GAL file DKR439NI.gal. Figure 8.5: Manually updated weights file ………………………. Figure 8.6 shows the histogram of the number of neighbours for the updated weights file DKR439NI.gal. Figure 8.4: Histogram of number of neighbours without “islands” 34 The linking applies not only to maps and tables, but also to statistical graphs like histograms. If you want to know where the two districts with 11 neighbours are located, click on the second right most bar in the histogram. Figure 8.5 depicts that the selected bar turns to yellow. Figure 8.5: Selection of the second right most bar in the histogram Because of linking, the location of the two districts with 11 neighbours can be seen in the map (Figure 8.6). Figure 8.6: Districts with 11 neighbours in linked map 35 Figure 8.7 shows that the districts in question are Bayreuth and Mettmann. Figure 8.7: Districts with 11 neighbours in the linked table Exercises 8.1 Check the weight properties for the weights file region5_ppt.gal! 8.2 Select the right most bar in the histogram for the weights file region5_ppt.gal and interpret the linked map and table! 8.2 Select the left most bar in the histogram for the weights file region5_ppt.gal and interpret the linked map and table!