Six Principles of Analytical Design
- by Edward Rolf Tufte
Critical thinking skills must occupy the top shelf of any aspiring data analyst’s arsenal. These include analyzing the data, interpreting, and presenting the data relevant and easily consumable by the reader.
Edward Rolf Tufte, in his book “Beautiful Evidence,” presents us with six principles for giving the data in an informative way.
The six principles are:
- Showing Comparisons
- Showing causality, mechanism, explanation, systematic structure
- Performing Multivariate Analysis
- Integrating Evidence
- Describing the Evidence (Documentation)
- Content is King!
In his book, Edward Rolf Tufte uses Minard’s map of Russia's French invasion, created in 1869, to beautifully explain all these principles. Interested readers can check this book.
Here, I am going to implement these principles with examples using Tableau. I will be using multiple datasets of the Air Quality Index of India from 2015 to 2020. I am using this data to see the Impact of COVID-19 on pollution in India. The datasets were downloaded from here. I have created a storyboard on Tableau on the Impact of Covid-19 on India’s Air Quality Index; please look at it here.
Indian Cities have always ranked top in the list of top polluted cities. But there was a drastic change in the air quality index after lockdown.
According to my analysis Ahmedabad, Delhi, Lucknow, Patna are the top 4 cities with the highest Air Quality Index. This data is until July 2020; right now, Delhi is the most populated city in India. We move forward with these four cities to see the impact of COVID-19 on their Air Quality Index. Let’s dive into our first principal.
Showing Comparisons
Edward Rolf Tufte says :
The fundamental analytical act is statistical reasoning is to answer the question “Compared with what” .
As whenever we are analyzing or providing evidence to our hypothesis, it is vital to make informed comparisons. Below we have a bar chart of the cities with their AQI from April (after the lockdown). The Air Quality Index for each city is around 3–4k, but we need to know this AQI is compared with what?
We have a hypothesis stating that the Air Quality Index has reduced drastically after COVID-19; it is essential to compare this with the AQI before April 2020.
Therefore, below is a bar graph comparing the state of AQI before and after COVID-19. We can now clearly see how the AQI has decreased drastically after April 2020.
Below is a bar chart race of different cities with their AQI from 2015 to 2020.
Furthermore, I compared the Air Quality Index using choropleth maps, making it easier to compare each city before and after lockdown. You can look at the interactive maps here.
Showing causality, mechanism, explanation, systematic structure
Above we answered the question of comparing with what?. We need to show the causal framework, mechanism, explanation, and systematic structure of what caused the AQI to reduce?. Due to the COVID-19 lockdown, the traffic decreased drastically, and the air was fresher. We can show this in a graph of the pollutants that contributes to the Air Quality Index. For now, I am just showing the air pollutants for Delhi.
We can see that the air pollutants has also reduced after the lockdown.
This graph also shows a noticeable pattern in these pollutants — the intensity of these pollutants increases during the winter season where people use an excess amount of biogas. Therefore while showing the difference between 2019 and 2020, I took only the first six months so that our analysis is not biased.
Performing Multivariate Analysis
Edward Rolf Tufte says :
Nearly all the interesting worlds(physical, biological, imaginary, human) we seek to understand are inevitably multivariate in nature.
We live in a multivariate world. Using multivariate analysis, we can tell a much deeper story.
I have plotted a graph showing a positive linear relationship between the Air Quality Index and air pollutants; I have also compared the data for the first two quarters of 2019 and 2020. We can see that after the 2nd quarter of 2020, the AQI is below 500.
Integrating Evidence
While we are presenting our report, our supporting evidence to the question could be of any form. As Edward Rolf Tufte says, what matters is the evidence, not the mode of proof. We can include evidence in the record of pictures, videos, words until and unless it provides answers to our question.
Evidence: These are two papers that support our analysis.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7299871/
Describing the Evidence
After including the evidence, the most important thing would be to describe each evidence understandably. This starts from describing our problem until we explain each of our evidence with proper title and content. You can find the storyboard here.
Content is King!
Above all these principles, to make a useful analysis, we need to take care of the content's relevance, quality, and integrity. The study should deliver the message in a way that is easier, impactful, and at the same time engaging for our audience.
References
[1]. Tufte, E., 2006. Beautiful Evidence. Available at:<www.edwardtufte.com> [Accessed 1 December 2020].
[2]. Peng, R., 2020. 5 Principles Of Analytic Graphics | Exploratory Data Analysis With R. [online] Bookdown.org. Available at: <https://bookdown.org/rdpeng/exdata/principles-of-analytic-graphics.html> [Accessed 1 December 2020].
