From CS 294-10 Visualization Sp10

Jump to: navigation, search


Original Question

My original question was to determine if there was a relationship between the number of PCs in a country, the number of mobile phones in a country and the GDP per person.

Data Format & Transformations

To answer this question, I collected a dataset of countries, the total number of PCs in the country, the number of mobile phone subscribers, the GDP and the population in 2004. I chose 2004 because it was the most recent year that had all the required information. I got my data from http://www.nationmaster.com. The data is a table format on the webpage, which I copied/pasted and formatted in Excel. Each of the quantitative measurements came in their own table, with a different set of countries.

To put all the data in one table, I had to join the tables on the common set of countries. I wrote a script in PHP to do this. Then, my data was ready for Tableau.


Iteration 1

On my first iteration, I decided to graph my data using a scatterplot with Number of Mobile Phone Subscribers on one axis and Number of PCs on the other. I used the size of each point to encode the GDP per person in each country. To spread out the data, I used logarithmic scales for both axes.

A general trend was apparent that the countries with larger GDPs had lower cell phone to PC ratios than those with higher GDPs.

Iteration 2

That was when I decided I wanted to add more data into the visualization to answer another question: How did literacy of a country affect the cell phone to PC ratio? To answer that question, I found literacy information for all the countries. I decided to use hue to encode literacy rates, resulting in the following graph.

Iteration 3

After seeing the result, I realized that the literacy numbers were all clustered around several rates, resulting in not much variation on the graph and not much information in terms of trends. So, I decided to remove that data point. Then, I decided that the graph was too cluttered with many countries, making it hard to read. So, I thought that separating by continents might be help to simplify the data. So, I found continent information for all the countries and used color to encode it. The resulting graph is shown below. I soon realized that separating by continent wasn't revealing any trends either.

Iteration 4

That was when I decided that maybe the actual values of Number of PCs and Mobile Phone Subscribers were making the graph complicated, and that only graphing the ratio against GDP per person would be useful. To graph this, I decided to use a bar char with countries on the vertical axis. The height of each bar represents the GDP per country. Color still represents the continent the country belongs to. The resulting graph is shown below.

Iteration 5

Looking at this graph, I realized that we had lost a lot of information by representing only the ratio, and that the current graph was misleading. However, it was much cleaner. So, then I tried to find a way to combine the data in the scatterplot with the simplicity of a bar graph. Then, I decided to use a stacked bar chart, where the two colors on the bar would be the Total Number of PCs and the Total Number of Phones (sorted by Number of PCs), and the height of the bar would encode GDP per person. The resulting graph is shown below.

Iteration 6

This is when I noticed an interesting trend in this visualization: when 2 countries have the same number of PCs, the country with the lower GDP per person generally had many more mobile phones. Because of the colors due to the stacked bar chart, I couldn't encode continent with a color. However, because I wanted to know the relationship between number of cell phone subscribers, number of PCs, GDP and population at a continent level, I decided to add a separate graph that showed that. My final visualization is shown below.

Final Visualization

The outer graph in this visualization shows the relationship between countries, the Number of PCs, the Number of Mobile Phones and GDP per Person. The inner graph shows the relationship between continents, the Number of PCs, the Number of Mobile Phones and GDP per Person. This does graph does answer my original question. It shows that when 2 countries or continents have the same number of PCs, the country with the lower GDP per Person (LowGDPCountry) has a higher number of phones than the country with the higher GDP per Person (HighGDPCountry).

Generally, the LowGDPCoutry has a much higher population, so the number of PCs per person would be really low. If the number of Mobile Phone Subscribers were to follow the same trend, then you would expect that the number of Mobile Phone Subscribers in both countries would be the same. However, because of the cost, portability and utility of mobile phones, we can see that they've taken off much more in countries with low GDP per person.

Paul Ivanov - Feb 22, 2010 05:31:09 pm

did you add the continents inset (inner graph) in your final image using Tableau, or was it just made it Tableau and then pasted in?

[add comment]
Personal tools