1 - Critique the visualization
Figure 1: The Original Visualization (with Markups)
1.1 Clarity
| S/N | Comments | Suggested Improvements |
|---|---|---|
| C1 | The neutral values of the likert scale on the diverging stacked chart are not centred on zero. This makes it hard for the users to compare the ratings across different categories and also makes it hard to tell whether the responses were skewed towards agreement or disagreement. | Redo the diverging stacked bar chart with the neutral values of the likert scale centred around zero. |
| C2 | The categories in the current chart is ordered alphabetically. This is not meaningful as there is no analytical findings nor message conveyed through unordered categories such as country names. | Sort the categories (rows) of the table by the proportion of a chosen response type and synchronize the order across both visualizations. This is important in order to compare the visualization across both charts. |
| C3 | Information is missing on how many respondents are there per country. This can be very valuable in providing a context to readers. | Add in the count of respondents from each country as a tooltip. |
| C4 | The respondents of each country is made up from different social strata. Each group can have different attitudes towards vaccination. By grouping them all by country, we run the risk of over-generalization. | Add in options for users to filter and slice the visualizations. This allows the user to glean more insight. e.g. comparing the willingness of respondents aged over 65 and unemployed in two different countries. |
| C5 | The second chart on the right shows the proportion of strongly agree responses as a point estimate. That can be misleading as there is often uncertainty associated with such survey data. | Change the chart to a point estimate with error bars showing the confidence interval of the responses. |
| C6 | The survey consists of a lot of questions and the current visualization only shows the responses relating to one question. This undermines the usefulness of the survey. | Add in options to view the responses of other questions in the survey. |
1.2 Aesthetics
| S/N | Comments | Suggested Improvements |
|---|---|---|
| A1 | Colours on the chart are randomised and do not follow a logical pattern. | Replot the chart with colours on a red-blue spectrum. Red should be on one end of the spectrum to indicate ‘Strongly Disagree’ and Blue on the other end to indicate ‘Strongly Agree’ |
| A2 | Scale of the diverging stacked bar chart starts at 0% and ends at 100%. This will not be sufficient to display the full range of responses gathered on vaccination opinions. | Adjust scale of the bar and the values of the responses to allow for negative values. This will allow the negative opinions (4 and 5) that show disagreement to be accurately represented. It will also help to answer the question posed by the title:Which country is more pro-vaccine? i.e. Change from the responses from 1 to 5 to -2 to 2. |
| A3 | Legend not properly named and labeled. While readers can infer the feedback when given by the two extreme values of ‘Strongly Agree’ and ‘Strongly Disagree’, it would be clearer to properly label each and every one of them. | Rename the legend and specifically label each and every one of the different colors on the diverging bar chart. |
| A4 | Clear titles on both charts | Follow and recreate a similar aesthetic. |
2 - The suggested visualization
The proposed design is as follows.
Figure 2: Sketching the Suggested Visualizations
3 - The Visualization process
The Data Source used is from the Imperial College London YouGov Covid 19 Behaviour Tracker Data Hub. The files can be found here.
3.1 Data Preparation
Download all 14 required data files from github as CSV files
Load all 14 files into Tableau and study the file structure.
Join all 14 files in a union using Tableau
Figure 3: Creating a union across all files.
- Create aliases so that the values are country names instead of countrynames.csv
Figure 4: Aliases for file names
- Pull required fields into the rows and column shelves.
- Rows : Table Name, RecordNo
- columns : disability, gender, House rooms, Household size, Employment Status, Household Children, Age, vac_1, vac_3, Vac2_1, vac2_2, vac2_3, vac2_6
- Create a view of the data for export
Figure 5: View data
- Export ALL data to a compiled file
Figure 6: Exporting a compiled file
8a. In excel replace values for in the fields ‘vac_1’, ‘vac_3’, ‘Vac2_1’, ‘vac2_2’, ‘vac2_3’, ‘vac2_6’. - ‘1 - Strongly Agree’ with ‘2’ - ‘2’ with ‘1’ - ‘3’ with ‘0’ - ‘4’ with ‘-1’ - ‘5 - Strongly Disagree’ with ‘-2’
Figure 7: Replacing values in excel
8b. This step can also be done by using aliases in Tableau. But a set of aliases will need to be created for each of the fields.
Figure 8: Replacing values in Tableau
- Change the data type for All vac columns to whole numbers
Figure 9: Changing data types
- Rename the field “Table Name” to “Country”
Figure 10: Renaming to Country
3.2 Steps taken for first Visualization
Create a new parameter showing the different questions given to the survey respondents. Match each of ‘vac_1’, ‘vac_3’, ‘vac2_1’, ‘vac2_2’, ‘vac2_3’, ‘vac2_6’ to their respective questions.
Expose the parameter by checking the ‘Show Parameter’ option.
Figure 11: Creating a new parameter for the survey questions
- Next, create a calculated field ‘Selected’
Figure 12: Creating a calculated field - Selected
- Next, create a calculated field ‘Number of Records’
Figure 13: Creating a calculated field - Number of Records
- Next, create a calculated field ‘Total Count’
Figure 14: Creating a calculated field - Total Count
- Next, create a calculated field ‘Count Negative’
Figure 15: Creating a calculated field - Count Negative
- Next, create a calculated field ‘Gantt Percent’
Figure 16: Creating a calculated field - Gantt Percent
- Next, create a calculated field ‘Percentage’
Figure 17: Creating a calculated field - Percentage
- Drag ‘Gantt Percent’ and ‘Country’ in the columns and rows shelves respectively
Figure 18: Populating the columns and rows
- Drag following fields into the marks card.
- ‘Selected’ to colour and change to Discrete Dimension
- ‘Percentage’ to size
- ‘Total Count’ to tooltip and change to Discrete
- ‘Selected’ to Detail and change to Discrete Dimension
- change the mark type from automatic to Gantt Bar
Figure 19: Populating the Marks card
- Change both ‘Percentage’ and ‘Gantt Percentage’ to compute using ‘Selected’
Figure 20: Changing the computation methods
- Filter away null responses in ‘Selected’
Figure 21: Filtering nulls
- Rename legend from ‘Selected’ to ‘Responses’ and assign a more logical colour scheme
Figure 22: Formatting the legend
- Create aliases for the numbered responses
- ‘-2’ to ‘Strongly Disagree’
- ‘-1’ to ‘Disagree’
- ‘0’ to ‘Neither Agree nor Disagree’
- ‘1’ to ‘Agree’
- ‘2’ to ‘Strongly Agree’
Figure 23: Creating Aliases for the responses
3.2 Steps taken for second Visualization
- Create a new parameter showing the response types given by the survey respondents. Match each of the numerical values to the responses given. Expose parameter by clicking on the Show parameter option.
Figure 24: Creating a new parameter for the survey responses
- Create calculated fields needed for 95% confidence interval. For starters - Create Z-Value for 95% confidence interval, ‘Z_95%’
Figure 25: Creating a calculated field - 95% Z value
28.Next, create a calculated field ‘Count Response Type’ which counts the number of response of the selected response type given by the parameter ‘Response Type’
Figure 26: Creating a calculated field - Count Response Type
- Next, create a calculated field ‘Prop’: proportion of response type compared to all the responses received
Figure 27: Creating a calculated field - Prop
- Next, create a calculated field ‘Prop_SE’ standard error of the proportion
Figure 28: Creating a calculated field - Prop_SE
- Next, create two calculated fields ‘Prop_Lower Limit 95%’ and ‘Prop_Upper Limit 95%’ to construct the error bars
Figure 29: Creating a calculated field - Prop_Lower Limit 95%, Prop_Upper Limit 95%
- Drag ‘Country’ and ‘Prop’ onto the canvas to create a table.
Figure 30: Creating a table with Country and Prop
- Change mark type to circle
Figure 31: Changing mark type to circles
- Drag all ‘Measure Values’ into the Marks Card
Figure 32: Moving Measure Values into Marks
- Drop all other ‘Measure Values’ from the Marks Card except ‘Prop_Lower Limit 95%’ and ‘Prop_Upper Limit 95%’
Figure 33: Moving Measure Values into Marks
- Drag ‘Measure Values’ to columns shelf.
Figure 34: Moving Measure Values into columns
Change resulting charts into Dual Axis
Synchronize both axes
Figure 35: Creating a Dual Axis chartt
- Change Mark Type to line
Figure 36: Changing Mark type to line
- Drag ‘Measure Names’ into both colour and Path for the Measure Values Card
Figure 37: Using Measure Names on Colour and Path
- Select colours for error bar and circle
Figure 38: Colouring the error bars chart
- Completed view of second visualization
Figure 39: Proportions with Error Bars
3.4 Adding Filters
- Create a calculated field ‘Age Groups’ by binning the age of respondents into bins of 10 years.
Figure 40: Creating Calculated Field - Age Groups
- Create a calculated field ‘Household Children (Groups)’ by binning the age of respondents into bins of size 2.
Figure 41: Creating Calculated Field - Household Children (Groups)
- Create a calculated field ‘Household Size (Groups)’ by binning the age of respondents into bins of size 3.
Figure 42: Creating Calculated Field - Household Size (Groups)
- Add all of the following fields to the filters card:
- ‘Age Groups’
- ‘Household Children (Groups)’
- ‘Household Size (Groups)’
- ‘Employment Status’
- ‘Gender’
Figure 43: Creating Calculated Field - Household Size (Groups)
- For all filters, select option to ‘All using related data sources’ to allow filtering across different visuals
Figure 44: Creating Calculated Field - Household Size (Groups)
3.5 Creating the dashboard
- For both the visualizations change the sort order to sort by ‘Prop’ so that the sorts will synch for both charts.
Figure 45: Changing Sort Orders for both charts
- Create a new sheet to create a title for the dashboard. This is not done on each individual worksheet because it will otherwise repeat the lengthy question. This allows the selected question to show up on the dashboard title. The title will be “Question: <Parameters.Survey Question>”
Figure 46: Creating a title for the dashboard
- Change title of first visualization to “Survey Responses by Country”
Figure 47: Renaming first visualization
- Change title of second visualization to “Proportion <Parameters.Response Type> to:”
Figure 48: Renaming second visualization
- Change the size of the dashboad to automatic. This is needed as the dashboard is expected to be wider and this will allow the dashboard to accomodate the two visualizations side by side. The wider visualizations will also show the proportions and error bar more clearly.
Figure 49: Changing size of dashboard to automatic
- Drag the main elements of the dashboard into a side by side arrangement
Figure 50: Initial layout of dashboard
- Add Reference line at 0 to the diverging bar plot
Figure 51: Adding reference line at 0
- Remove country names from the first visualization as both the visualizations will have the same sort order and can share the labels in the middle.
Figure 52: Removing Country Names for more Space
Rearrange filters neatly.
Add in reference to data source.
Turn on animations for smoother transitions while filtering
Dashboard is complete.
4 - The improved data visualization
The completed visualization is as follows.
Figure 53: Snapshot of the completed dashboard
The interactive dashboard is uploaded onto Tableau Public Server and can be found here.
5 - New Insights
5.1 Vaccine avoidance due to side effects
Using the orange boxes - In the survey question: “If a Covid-19 Vaccine were made available to me this week, I would definitely get it.” We can see that a large proportion of the French respondents disagreed strongly (ranked 1st when sorting by proportion ‘Strongly Disagree’). When comparing this against the results of the question: “I am worried about the potential side effects of a Covid-19 vaccine.” The French respondents ranked top by proportion of ‘Strongly Agree’ responses.
Conversely, using the green boxes, we can see the countries of Netherlands and United Kingdom having correspondingly high levels of vaccine acceptance and low levels of worries about side-effects.
Figure 54: Resistance to vaccine could come from fear
5.2 Bigger familes have mixed views on vaccines
On the same survey question: “If a Covid-19 Vaccine were made available to me this week, I would definitely get it.” You can see a huge standard error in the proportion of ‘Strongly Agree’ responses. This could be due to a couple of reasons. The first could be due to the actual divergence in views in large families. The second is could be due to a smaller sample size, which is the case. (Most of the survey respondents had family sizes of 1-3).
A takeaway on this is to view the data as a whole without filtering on household size unless the number of survey responses from larger households can be collected.
Figure 55: Increased SE in bigger families
5.3 Many are waiting before taking vaccine
Following the same line of thought from the previous insight. Vaccine worries tend to ease off with time with better research and testing. The general attitude towards vaccine warms up considerably when vaccines were to be made available in one year instead of one week. This can be seen in lower number of disagreeing responses across the board (orange boxes) and higher number of strongly agree across all countries (green boxes).
A takeaway for the authorities could be that while aggressively ramping up vaccination capacity, they need to work on converting opinions on vaccine safety as well. Ambitious targets to vaccinate whole populations in months could be futile if the population is resistant towards taking the vaccines.
Figure 56: Vaccine opinions improve over time