Keywords: Corona, COVID-19, Chinese Outbreak

Since the Corona / COVID19 outbreak last December 2019 that started in Wuhan, China - a lot has been written about the phenomenon and it was confirmed as pandemic by World Health Organization on the 11th of March 2020. According to Johns Hopkins University 184 countries (at the time of writing this post) have.been affected by the virus and have registered cases of COVID 19. 

From a statistical data science perspective pandemics are interesting subjects to study, as they cover a lot of key statistical concepts that we use in data science. Eg distribution theories, significance levels etc. My objective in this series of posts is to try and balance a simplification of what is going while also keeping it understandable for audiences that are less data-literate. Implying that it is preferable to keep research questions small and by answering the individual questions per post - we are gradually building knowledge about the virus and it's outbreak and at the same time are aiming to help readers become more data-literate.  

For this series of posts I will be zooming in on the following research questions / topics: 

In this second post I will be having a closer look at how the virus developed in China? With the use of R (language and environment for statistical computing and graphics) I will be using data science technique and methods like descriptive statistics, geo-spatial analysis and multi-variate analysis. The insights can be used to track and analyse the spread of COVID-19 in China. 

Initially, while performing the analysis I used the data obtained from Johns Hopkins University's Github a place where they provide daily files with various information, e.g. number of new cases, number of death related to COVID-19 etc. This data is also forms the basis of their well-know dashboard that a lot of media use as main source of information in their articles. 

However halfway trough the analysis I learnt about the R-package called Coronavirus written by Rami Krispin. A wrapper that makes a data scientist life much easier, as it - in the background fetches the data from Johns Hopkins and there is no longer a need to have extensive code for obtaining and restructuring the data in my own code. Awesome! Very convenient as well is that longitude and latitude are also available features in that data. Meaning we can use these coordinates to plot the spread on a map.The images below shows the data available. The first image shows all available data and the second image you can see that for each province on a daily basis 3 'types' of records are presented, being the number of confirmed cases, nummer of deaths and the number of recovered patients  

Now that all the data is available in out R environment we can start by visually inspecting the data. Let's start with creating a histogram of the daily new confirmed cases for the entire Chinese Mainland. This might give us insights on how the virus is developing across the country. The figure below shows some volatility in the development of the virus. Something that we wouldn't expect based on a basic assumption in virology, namely that of exponential growth.What we should be looking for from a statistic perspective is a normal distribution (bell curve). This might imply that we need to add more detail to the histogram by taking the perspective of the different regions, to find these bellcurves. (ps also very noticeable is of course the spike seen on the 13 of February - let's investigate that on as well)

The figure below show the epidemic curves per chines province. Here we can see that the spike on a national level on 13 February is caused by the confirmed cases in Hubei. Being the province where Wuhan is located and the Corona/COVID-19 virus was fist discovered. 

Furthermore we seen that on a more detailed level in comparison to our first country histogram, that for provinces we discover more bell-shaped curves. Giving us hits towards predictability in a later stage. However it does need to be taken in account that we have a limited amount of days in our analysis set, and matching this to the assumption that the pandemic will have an exponential growth predicting the curve might be very tricky. Instead of focusing on predictability, let's focus on obtaining more insights on the spread.

An other important aspect in researching the spread is to identify epic centres of the spread. Meaning that we can have a look at the relative contribution of each state compared to the national figures. 

In order to identify the epic centres of the spread, different visualisation can be used. An option would be to add categories (being the provinces) to the histogram of the national figures. However keeping in mind that there are a total of 31 provinces this might become a bit messy. Therefore a treemap might be helpful with the identification of huge contributors to the national figures.Where the size of the category (in our case provinces) illustrates their contribution to the total amount of infections. 

In the treemap below can be seen the top 3 of infection ares are: (note that in the figure below the percentages are not displayed for readability purposes in terms of not showing all the 30 percentages)

1. Hubei: 82% of the infections.(+/- 58k of 70K). 
   
2. Guangdong:1,9% (+/- 1,3k of 70K).
   
3. Henan: 1,7% (+/- 1,2k of 70K).
   

A limitation of our investigation is the timeframe of our analysis set. The background of this limitation is that at the time of the analysis the Dutch spread just started to develop. From that perspective I wanted to keep groups comparable. Keeping in mind that my daily research endeavours were mainly based on statements in the media. Meaning that there was a fear that the virus may or is spreading as fast as it did in China. Therefor in this research I wanted to find support for such claims. For future research it may however be very useful to re-perform the analysis with more data and also look more at predictability of epidemic curves.

All the best,

Ditty