Transforming and Visualizing Data

Figure 4. The interactive Data Studio dashboard for investigating The Guardian's Data Records on the 2011 Riot Incidents in the United Kingdom (Link).

1. I renamed the “Authority” variable values after identifying and fixing spelling errors in the names of local authorities, referring to the official sources[68] with the lists of names of all authorities in the United Kingdom.
2. I removed the “Icon name” column, as The Guardian likely created it for mapping purposes, visually defining the data points (1 point as “capital_small,” 243 points as “placemark_circle,” and 1 point as “placemark_circle_highlight”). As such a variable is not coming from the primary source data I focus on, and as the documentation on the creation of such a variable is missing, I will not be using it for the purposes of visualizing data in this assignment.
3. I substituted empty cells in the “Source” and “Picture links” columns with a string value of “Missing Data”, to emphasize the records that lack credibility, as The Guardian refers to the dataset as including “every verified incident”[69] of 2011 riots, though its entries are missing the “Source” value to verify them.
4. I geocoded the “Location details” variable values by using the extension “Geocode by Awesome Table,” which extended the dataset with corresponding “Latitude” and “Longitude” values. Such transformation was necessary to represent the data geospatially in the dashboard visualization. During this process, 1 out of 245 entries geocoded incorrectly (“Location details” with “Chatham, Kent”) (placing the coordinates in Canada!); I resolved this issue by also adding the country name in the cell, and thus 245 out of 245 entries were geocoded successfully.
5. As two other columns of “Time” and “Authority” had empty cells, I substituted them with a “Missing Data” string value as well, to transparently communicate that such data is missing for some entries, which should be accounted for and emphasized when making inferences about the data in this source.
6. I renamed some of the original variables to avoid running into ingestion data errors in Google’s Data Studio and BigQuery, as these tools require ingesting variables with a one-word name length. I also renamed other variables to follow a more clear and concise naming convention. In the table below I document the updates made to the original source variables, a rationale for such updates, and variable descriptions:

Motivating the Dashboard Data Visualization

• Lacking Evidence: 2 of the data records are not supported with any evidence (in “Source1”), though they are presented in a dataset along with more credible, evidence-based records (with the links to articles, police records, etc.). The columns such as “PhotoLink”, “Authority”, “Timestamp” also have missing data.
• Lacking Geospatial Representation: Tabular representation of data does not effectively represent spatial differences in the records, which can be mapped and further explored by visualizing locations behind records, data clusters, etc.
• Lacking Representation of Change over Time: As records include dates from April 4 to April 10, 2011, we can benefit from looking at the change over time in all records, as well as the filtered records (e.g., by “Authority”), which we fail to grasp by only looking at a tabular representation of such data.

• Lacking Explorability on the Levels of Authority/Date/Time/Source: To explore different historical arguments pertinent to the data, it can be especially useful to analyze the records by filtering them on separate levels - looking into select local authorities, dates and timestamps, and the data sources.

• Conditional Formatting: The possibility of incorporating tabular, conditional formatting in a dashboard can visually emphasize missing data. While such formatting can be achieved via Google Spreadsheets,[70] dashboards allow to cross-filter of multiple visualizations, connecting a tabular representation of data with its corresponding map representation, charts, and other visualizations.[71] We can conditionally format the “Source1”, “PhotoLink,” as well as “Authority” and “Timestamp” values by flagging their missing record values in red.
• Geospatial Chart: The dashboard can include a map of data points, given that we have the values of Latitude and Longitude; when we select a data point on the map, we can display its information in other dashboard figures such as charts and tables (e.g., representing the record’s variables in a table format, as it might not be possible to visualize all of the available record’s data on a map).
• Time Series Chart: We can create a time series histogram that displays how the number of records change between August 4 and August 11, 2011; once we select one of the bars on the histogram, the other dashboard component can be cross-filtered - e.g., the geospatial chart will only map the records with the exact value of date selected from a histogram.
• Filtering by: To investigate the data on multiple variable levels, we can create filters for “Authority,” “Date,” “Time,” and “Source” variables, and once the records are customized based on these filters, the proposed visualizations above will display only the data that is filtered by select values of these variables.

Figure 5. The interactive Data Studio dashboard landing page of the Capstone Project (Link).

Dashboard Documentation

Figure 6. Page 1 of dashboard documentation including description of the data source, a way to connect it to Data Studio, and steps to produce Filters and Figure 1.

1. To filter data in the dashboard by Authority, Date, Time, and Source, we select "Add a control" and "Drop-down list."
2. In the "Setup" section, we input a corresponding filter variable in the "Control field"; for "Metric," we choose "Record Count," marking "Show values," and for "Order," we select "Dimension" and "Ascending" to display the values alphabetically or chronologically.
3. In the "Style" section, we select the "Enable search box" to search for records of interest inside the filter boxes.

Figure 7. Page 2 of dashboard documentation including steps to produce Figures 2-3.

1. We click on "Add a chart" and select the "Bubble Map" under the "Google Maps" section.
2. In the "Setup" section, under "Fields," we input the "LatLon" location variable, which we compute by merging the dataset's Latitude and Longitude fields; under "Tooltip," we input the variable "Place," which will be shown once you click on the data points on the map; under "Chart interactions" we enable "Cross-filtering," which will sync actions of the user on the map to other visualizations - displaying the selected data point on the histogram of Figure 2 and table of Figure 3.
3. In the "Style" section, we select "Map" under the "Background Layer" field, define "Style" as "Dark," and "Layer Type" as "Bubbles"; under "Map Controls," we mark "Allow pan and zoom" to be able to zoom in and out on the map.

1. We click on "Add a chart" and select the "Time Series Chart" under the "Time Series" section.
2. In the "Setup" section, under "Dimension," we input the "Date" variable; we select "Count" under "Metric" and enable "Cross-filtering" for this figure, too - once a user clicks on the bar of the histogram (a specific count of records per date), the map will zoom in on the corresponding location automatically, both map and the table will display the records behind the selected histogram bar.
3. In the "Style" section, we select "Bars" under "Series #1" and mark "Show data labels"; under "Axes," we mark "Show axes"; and under both "Left Y-Axis" and "X-Axis" we select "Show Axis Title."

1. We click on "Add a chart" and select "Table" under the "Table" section.
2. In the "Setup" section, we input variables of "Date," "Time," "Authority," "Place," "Description, "Source," and "Photo"; we choose "Ascending" under both "Sort" (input "Date") and "Secondary sort" (input "Time" to order the records chronologically; we enable "Cross-filtering" to allow to click on the table entry and reflect its data spatially on the map.
3. In the "Style" section, we mark "Show Header" and "Wrap Text" under "Table Header" and mark "Row Numbers" and "Wrap Text" under "Table Body."
4. In the "Conditional Formatting" under "Style," we define 4 rules to emphasize the missing data in the "Source," "Photo," "Time," and "Authority" variables:

Highlighting Dashboard Functionality

Figure 9. View of the dashboard when a data point on the map is selected.

Considerations for Dashboard Improvements

Figure 11. View of the dashboard’s page 3,  representing the Home Office’s official data on riot arrests, and Newburn et al. interview data records on the 2011 Riots in the United Kingdom.