1.0 Critique of Visualisation
The original visualisation can be seen below.
1.1 Clarity
The use of stacked bar charts results in readers having to estimate the proportion across different countries. This is hard to compare, as the baseline of each option keeps shifting, which may lead to misinterpretations.
Both charts are sorted differently, with Chart 1 (general response chart) being sorted alphabetically by country, while Chart 2 is sorted by percentage of strongly agrees. As they are put side-by-side, users may misinterpret that the Y-axis categories are the same.
Chart 2 has many values that are quite close, such as Sweden, Canada, Norway, Finland, and Italy. It is hard to make comparisons between these countries without any percentage values or reference lines shown.
The x-axis of Chart 2 is only up to 60%. This may confuse the reader as it would over-emphasize those that strongly agree since the max value for Chart 2 is the same length as one of the stacked bar charts in Chart 1.
1.2 Aesthetic
The colour selection of the bar charts is not optimal. Red is usually used as an eye-catching colour for poor performance or as a highlight. Here, it is used for Option 3, neutral. This is against convention.
The same blue is used for Option 1, Strongly agree, and the percentage of strongly agreed to vaccination. The same colours can be confusing at first glance.
There are typos in Chart 1 title, as vaccine is mis-spelled as “vacinne”
The units for the x-axes of both graphs are not fully standardised as Chart 1 has no decimal places, but Chart 2 is up to 1 decimal place.
2.0 Alternative Design
The proposed design is as follows.
2.1 Clarity
Instead of a stacked bar chart, a likert scale chart is used to represent the diverging nature of the responses. This more clearly represents the proportions and visual scale of the differing opinions of different countries.
As there is only 1 x-axis for the likert chart, there is no confusion on the size and scale of different bars. X-axis is also fixed to prevent shifting of the 0 value.
The values for the uncertainty chart follows the statement being mentioned in the likert chart, so it is clear that both charts are talking about the same statement / question.
The reference band for different confidence levels will be shaded slightly less dark compared to the mean value, with the confidence level being stated clearly to show uncertainty.
Both visualisations are sorted in the same way, so as to allow for easier comparison across visualisations.
2.2 Aesthetic
The colours selected will be more in line with convention, with redder colours representing negative opinions.
The legend is updated to reflect the options for opinions more clearly by describing all options (1 = Strongly Agree, 2 = Agree, 3 = Neutral, 4 = Disagree, 5 = Strongly Disagree)
A reference line at 0% is used in the likert chart to further emphasize the diverging opinions.
Different parameters are used to enable the user to select their preferred category for further analysis instead of having an aggregated data visualisation, as there might be variation of opinion between groups.
3.0 Proposed Visualisation
Please view the interactive visualisation on Tableau Public here.
4.0 Step-by-step Guide
4.1 Graphing a Likert Chart
| No | Step | Action |
|---|---|---|
| 1 | In the data source tab, select the New Union function and import all data sheets using a manual connection. |  |
| 2 | Export the data using the View Data function. |  |
| 3 | Import cleaned data into a new Tableau sheet. Split Table Names using the custom split function to create a new column called “Country”. |  |
| 4 | Create a new parameter field for question selection. Use the list function and list out the relevant questions as variables. Key in text fields in the Display As column to explain the questions for more clarity. |  |
| 5 | Create new Calculated Field and set the cases when different questions are selected. |  |
| 6 | Create new Calculated field to convert them from string into an integer format. |  |
| 7 | Create a record list using a calculated field and setting the value as 1. |  |
| 8 | Create a Total Count Calculated Field. |  |
| 9 | Create a Count Negative Calculated Field to define the values for the likert scale diagram. |  |
| 10 | Create a Total Count Negative Calculated Field to consolidate all negative values. |  |
| 11 | Create a Gantt Start Calculated Field to define the start of the Gantt Chart. |  |
| 12 | Create a Gantt Percentage Calculated Field to define the size and scale of the likert scale diagram. |  |
| 13 | Create a Percentage Calculated Field to define the size of the bar chart. |  |
| 14 | Insert variables into the Rows and Columns selector. |  |
| 15 | Add Question selector to Detail and Colour, Percentage to Size, and change the graph type to Gantt Bar. |  |
| 16 | Ensure that the Gantt Percentage chart is calculated using the Question Selector Variable. |  |
| 17 | Add a reference line at the 0 value on the x-axis. |  |
| 18 | Base visualisation with variable selections for question looks like this. |  |
| 19 | Create another parameter for demographic selection, this time filling in the values for the selected variables. Change the display name as required. Use “(All)” in the parameter field to ensure that there is a master parameter selection. |  |
| 20 | Create a Calculated Field that sets the correct demographic case when the associated variable is selected. |  |
| 21 | Ensure that all the demographics are in string / categorical format. |  |
| 22 | Create groups for the Age variable to bin the values to make them into categorical variables. |  |
| 23 | Bin sizes should be something reasonable, for example, a range of 5 years. |  |
| 24 | Repeat for the group creation step for Household Children, grouping them into fewer categories. |  |
| 25 | Repeat for the group creation step for Household size, grouping them into fewer categories. |  |
| 26 | Display the filter for the Demographic Selector Calculated Field. |  |
| 27 | Add a filter to exclude Null replies for the Question Selector variable if any. |  |
| 28 | Add aliases to the Question Selector variable to give more clarity for the values. |  |
| 29 | Confirm that the Question Selector is sorted correctly. If there are errors, select on an option and adjust with the arrows. |  |
| 30 | Add the Demographic Selector to the Rows, and the base design of the sheet is created. |  |
| 31 | Fix the x-axis to prevent the reference line from shifting when parameters change to avoid confusing users. |  |
| 32 | Edit the tooltip to give explanations on the various values and parameters used. |  |
| 33 | Right click and rename titles to more suitable statements for dashboarding. |  |
| 34 | Rearrange panels to suit aesthetic preference and add captions if needed. This is a view of the final product. |  |
4.2 Graphing Uncertainty
| No | Step | Action |
|---|---|---|
| 1 | Create a Calculated Field called Proportion with the following formula. This is to find the number of respondents that gave a Strongly Agree or Agree score. |  |
| 2 | Create a Calculated Field called Prop_SE with the following formula. This is to find the standard error for the uncertainty of the proportion. |  |
| 3 | Create Calculated Fields for z-values for various confidence levels. In this case, z-values for 90%, 95%, and 99% were created. |    |
| 4 | Create a Parameter to select desired confidence level. Keep data types as String and list out the various confidence levels that were created in the previous step. |  |
| 5 | Create a Calculated Field to select for the confidence level input with the following formula |  |
| 6 | Create a Calculated Field to determine the correct lower bound of the confidence interval, using this formula. |  |
| 7 | Duplicate the Calculated Field and change the formula to create the upper bound for the confidence interval. |  |
| 8 | Drag Proportion and Country into the Columns and Rows, respectively. This will create a preliminary baseline for the uncertainty chart. |  |
| 9 | Drag Measure Values into the Columns tab and remove all irrelevant information, leaving only Prop_Lower Bound and Prop_Upper Bound. |  |
| 10 | Change the chart to a line graph, then drag Measure Names into the Detail Tab. Change Measure names from a Detail to a Path. |  |
| 11 | Create a dual axis chart by right-clicking on the x-axis. Remember to select Synchronise Axis to align the x-axes. |  |
| 12 | Repeat step X to Y for a comparison level of confidence interval. Drag these new upper bounds and lower bounds into Measure Values. |  |
| 13 | Ctrl-drag Measure Names into the Colour tab. |  |
| 14 | Select the Colours tab, edit the various colours so that the Lower and Upper bound match up with the correct pair. |  |
| 15 | Click on the arrow at the side and select sort. Pair up each of the respective lower and upper bounds. |  |
| 16 | Drag Demographic Select field that was previously created into the row to allow for variable selection for the demographic parameters. |  |
| 17 | Edit the tooltip to display information to the user. |  |
| 18 | The final layout of the uncertainty sheet is complete. |  |
4.3 Dashboarding
| No | Step | Action |
|---|---|---|
| 1 | Drag the Uncertainty sheet into the dashboard. Rearrange and align to fit the screen and ensure readability. |  |
| 2 | Select the arrow for the Country Parameter, go to Apply to Worksheets sub-menu and ensure that the All Using This Data Source option is selected. This will ensure that this Parameter can control both visualisations. Repeat step for other relevant parameters that have been included. |  |
| 3 | The final dashboard is complete. |  |
5.0 Derived Insights
- France is the most “anti-vaccine” country, with the least confidence in a vaccine that they can take next week, with almost 40% of participants expressing a Strongly Disagree opinion for taking a vaccine next week. It also has the smallest proportion of people agreeing with the statement, at only 39%.
This has been supported by other surveys conducted by other agencies and polling companies. There may be more instances of vaccine-skepticism or other forms of misinformation driving this in France. This may also be due to cultural norms whereby France is generally more skeptical of vaccines. Euronews, 2021
- Japan is the most afraid of side effects from a vaccine, with 79.2% of respondents expressing a Strongly Agree or Agree opinion on worries for side effects for a vaccine. Singapore follows closely behind with 75.8% of participants agreeing with the comment. Generally, all 3 Asian countries rank relatively high on this statement, with all 3 of them placing in the top 5 countries being most concerned. France has also ranked quite high, most possibly being linked to being most skeptical of a vaccine.
Past vaccine development usually takes many years to develop, and that has caused some concerns, especially in Japan, which may have contributed to the lower confidence in vaccines and the greater concern with side effects, as shown in the visualisation. (Japan Times, 2020)
- Males in general are more willing to take a vaccine this week as compared to females. This may be linked to the general worries about side effects for a vaccine, as females are more likely to be worried about side effects, which may in turn decrease their propensity to receive a vaccine.
- Singapore has the most confidence in the government, as the proportion of Strongly Agree and Agree is the highest when asked about whether the government will provide an effective vaccine. This might be due to general high confidence levels in Singapore for the government due to its past records of steering Singapore out of crises.
France is the least confident in its government, possibly due to its general lack of confidence in effective vaccines.
- Singapore ranks quite low when asked if there would be any regrets for not taking the vaccine. This might be potentially due to Singapore’s virus situation being well controlled, with only a few cases each day (at point of writing) compared to the other countries on the list. As such, the fear of side effects may outweigh the benefits of immunity in some respondents.
In comparison, the UK, with the highest death toll in the world per capita (Neilson, 2021), would regret the most.
