DataViz Makeover 3

DataViz Makeover

Dive deeper into the spatio-temporal patterns of armed conflict in selected South-east Asia countries between 2010 - 2020.

Author

Affiliation

Bai Xinyue

 

Published

March 19, 2021

DOI

Data Visualisation Link (Tableau Public) - https://public.tableau.com/profile/xinyue.bai#!/vizhome/dataviz_16162565365570/South-eastAsiaArmedConflictAnalysis?publish=yes

1. Background Information

For this visualisation makeover, I have used data from ACLED. ACLED focuses on tracking a range of violent and non-violent actions by political agents, including governments, rebels, militias, identity groups, political parties, external actors, rioters, protesters and civilians. Data collected is used regularly to study trends in political violence and protest around the world. In particular, this post is interested in exploring the spatio-temporal patterns of armed conflict in selected South-east Asia countries, including Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Thailand and Vietnam, between 2015 and 2020.
In this blog, I will makeover visualisation on South-east Asia Armed Conflict provided by Prof.Kam, by looking into the following questions:

2. Critiques and Suggestions

original visualisation

Figure 1: original visualisation

2.1 Clarity

SN Critique Suggestion
1 Y-axis of the line chart does not start at 0, making it hard to compare between event types. Make y-axis start at 0 and make it consistent across all event types.
2 Also, the name of y-axis “Count of Sheet1” is unclear. Label the y-axis as “Count of Event Type”.
3 It’s redundant and hard to compare having 6 separate line graphs. Combine 6 graphs into 1 graph and use different colours to represent event types.

2.2 Aesthetics

SN Critique Suggestion
4 Points on the map are overlapping. It’s hard to distinguish the event type at each area.

Apart from a map of events’ exact location, we can have another map to summarise information of events:

  • Display events in a state/province level instead of precise location.
  • Use unclassified choropleth map to demonstrate the event density in each state/province.
  • Use pie chart to show the distribution of event type in the tooltip.
5 Order of the event type on the line chart is not aligned with the colour legend.
  • Make order of the line chart align with the colour legend.
  • Match the line colour with the colour legend.

    • 2.3 Interactivity

    SN Critique Suggestion
    6 Interaction techniques used in this visualisation are highlighting(event type) and filtering(country), allowing audiences to see the exact location of all events between 2010 - 2015 and the total counts of each event type over the years for every South-East Asia country. However, the interactivity is very limited. For example, what if audiences want to know the location of events in a specific year, as it might be different every year? We can add more filtering options, such as year, range of dates and event type.
    7 Since only two techniques are used in this case, there are other type of interactivity can be applied to get a more comprehensive understanding of the data. For example, “Overview first, then details-on-demand” from Shneiderman’s mantra, we can provide a map with an overview of the event frequency with respect to state/region and a pie chart for each state on detailed distribution of event type in the tooltip. Similarly, for each event location, we can include names of two conflict groups (Actor 1 & Actor 2) in the tooltip.

    3. Proposed Design

    3.1 Sketch

    Figure 2: sketch of proposed design

    3.2 Advantages of Proposed Design

    1. Proposed design allows more filtering options (e.g. year, range of event date, event type).
    2. Proposed line graph is more succinct and easier to compare between event types (6 lines in one plot). Also, adding information of fatalities offers audiences a better understanding of the severity of events.
    3. Proposed state-level density map offers a more comprehensive view of the data, by showing frequency and distribution of events.
    4. Proposed heatmap of event frequency between different groups allows audiences to know which two parties are frequently involved in an armed conflict. This information is very useful in preventing or mitigating future conflicts.

    4. Data Visualisation Step-by-Step

    4.1 Data Preparation

    1. Import Southeast-Asia_2010-2020_Oct31.xlsx file into tableau.

    2. Change the data type of Year to Date. Figure 3

    3. Select data between 2015 and 2020, by clicking Filters - Add from the top right-hand corner.

    1. Change the data type of Admin1 to Geographic Role -> State/Province.
      Figure 8
    2. Similarly, change the data type of Interaction to String.

    4.2 Creating Visualisation

    4.2.1 Spatial Analysis Dashboard

    1. Open a new worksheet (to create an event density map).
    2. Double click on Admin1 and drag Country to Detail. Figure 9
    3. Change graph representation from Automatic to Map under Marks. Figure 10
    4. Drag Event Type to Color and apply Measure -> count, to get total count of events for each state.
      Figure 11
    5. Drag Country, Event Type, Event Date, Year to Filters and click Show Filter.
    1. Change the color palatte to Red. Figure 14
    2. Change the style of map from Light to Normal (Map -> Map Layer -> Style -> Normal).
    1. Open another new worksheet (to create a pie chart of event distribution).
    2. Drag Admin1, Country to Detail, Event Type to Color, and change graph representation from Automatic to Pie under Marks.
    3. Drag Event Type to Angle, applying Measure -> count (shown in step 4), Quick Table Calculation -> Percent of Total and Compute Using -> Event Type.

    1. Similarly, drag Event Type to Label and apply the same procedure as in step 10. Final look of the pie chart:
      Figure 19

    2. Include the pie chart in the Tooltip of state-level event density map.

    1. Open another new worksheet (to create a map of event exact location).

    2. Drag Longitude and Latitude to Columns and Rows respectively, and change both variables from Measure(Average) to Dimension.
      Figure 22

    3. Change graph representation from Automatic to Shape under Marks and drag Event Type to Color and Shape, adjust the size. Figure 23

    4. Drag Actor1, Actor2, Interaction to Detail and change Interaction from Measure(Count) to Dimension (as shown in step 14), allowing audiences to know which party is involved in each conflict, by hovering over the points.
      Figure 58

    5. Change the style of map from Light to Normal (as shown in step 7).

    6. Open a new dashboard.

    7. Change the size of dashboard to Generic Desktop. Figure 24

    8. Drag 2 maps into dashboard and change representation of all filters excpet Event Date to Single Value (Dropdown).
      Figure 25

    9. Apply all filters on 2 maps, by adjusting Applying to Worksheets -> Selected Worksheets.

    1. Edit title of 2 maps. Figure 28
    2. Edit unknown locations, by googling online.
    1. Final look of Spatial Analysis Dashboard. Figure 31

    4.2.2 Temporal Analysis Dashboard

    1. Open a new worksheet (to create a line graph showing frequency of each event type over years).

    2. Drag Year and Event Type to Columns and Rows respectively, and change Event Type from Dimension to Measure(Count) (as shown in 4.2.1. step 14).

    3. Drag Event Type to Color.

    4. Drag Country, Event Date, Event Type to Filters and click Show Filter. Final look:
      Figure 32

    5. Open another new worksheet (to create a line graph showing count of fatalities by event type over years).

    6. Drag Year and Fatalities to Columns and Rows respectively and drag Event Type to Color. Final look:
      Figure 33

    7. Open another new worksheet (to create a heat map showing frequency of conflicts between two parties over years).

    8. Drag Year and Interaction to Columns and Rows respectively, and change Interaction from Measure(Count) to Dimension (as shown in 4.2.1. step 14).

    9. Drag Event Type to Color and change it from Dimension to Measure(Count) (as shown in 4.2.1. step 14).

    10. Change the color palatte to Red (as shown in 4.2.1. step 6). Final look:
      Figure 34

    11. Open a new dashboard.

    12. Change the size of dashboard to Generic Desktop (as shown in 4.2.1. step 19).

    13. Drag 1 heatmap and 2 line graphs into dashboard.

    14. Change representation of Event Type filter to Single Value (Dropdown) and representation of Country filter to Single Value (List) (as shown in 4.2.1. step 20).

    15. Apply all filters on 3 graphs, by adjusting Applying to Worksheets -> Selected Worksheets (as shown in 4.2.1. step 21).

    16. Edit title of 3 graphs (as shown in 4.2.1. step 22).

    17. Add a Text Object to dashboard, explaining the meaning of interaction code.
      Figure 35

    18. Final look of Temporal Analysis Dashboard. Figure 36

    4.2.3 Main Dashboard

    1. Download images of earth and calendar online, create a custom folder under My Tableau Repository -> Shapes and put images inside.
      Figure 39
    2. Create a calculated field named image, with value 0. Figure 40
    3. Open a new worksheet (to create a Button for Spatial Analysis Dashboard).
    4. Drag image to Detail and change graph representation from Automatic to Shape under Marks (as shown in 4.2.1. step 3).
    5. Change the default shape to the image of earth.
    1. Change the layout of the worksheet from Standard to Entire View.
      Figure 44
    2. Edit the content of Tooltip to “click to view the spatial analysis” Figure 45
    3. Similarly, open a new worksheet (to create a Button for Temporal Analysis Dashboard) and a new worksheet (to create a Button for Return to the Main Dashboard).
    4. Open a new dashboard and change the size of dashboard to Generic Desktop (as shown in 4.2.1. step 19).
    5. Drag 2 image worksheets into dashboard and add border. Figure 49
    6. Add a Text Object to dashboard (as shown in 4.2.2. step 17), giving a brief introduction of this data visualization.
    7. Add an Image Object to dashboard.
    1. Add 2 Text Object to dashboard as titles of spatial and temporal analysis dashboards, and change their background color.
      Figure 48
    2. Add interactivity on the Earth and Calendar worksheets.
    1. Final look of Main Dashboard: Figure 55
    2. Similarly, add interactive Return Button on Spatial Analysis Dashboard and Temporal Analysis Dashboard.

    5. Major Observations

    1. As we can see from the event density map, Shan province in Myanmar has the highest number of political conflicts between 2015-2020 and 63% of events in Shan are battles. By jumping to the Temporal Analysis Dashboard and select Shan, battles occurred most frequently in 2015, between state forces and rebel groups, resulting in more than 800 fatalities.
    1. As we can see from the temporal heat map, the number of protests (1-sided protest) has been increasing over the years. By looking into spatial dashboard, the majority of protests took place in Jakarta.
    1. Although the number of protests is relatively high compared to other type of events, fatalities of protests are generally low, whereas violence against civilians occurred frequently between 2015 - 2020 and its fatalities are also high, peaking at 3569 in 2016. Figure 65
    2. Armed conflicts between rebel groups and protestors, political militias and identity militias, identity militias and rioters, identity militias and external forces, protestors and civilians are very rare. Figure 66
    3. Philippine and Myanmar account for majority of the events occurring between 2015 and 2020, especially in battles and violence against civilian event types. In the events of riots, Indonesia makes up a larger proportion.

    Footnotes