A2-NicholasKong
From CS294-10 Visualization Fa08
Contents |
Picking a Domain
I had heard that Japan has an exceptionally high suicide rate. The causes of this are debatable: perhaps intense societal pressure to conform, or perhaps isolation. Whatever the true causes, it seems likely that suicide would be more of a problem in more developed nations as most of the preventable causes of death will have been eliminated. At least, this was my hypothesis, and I thought it would be interesting to explore this further.
Initial Question
As stated in the previous section, the question was simple: are suicide rates correlated with the level of development of a country?
The Dataset
The World Health Organization compiles statistics on causes of death around the world. The most substantial dataset containing a complete cause-of-death breakdown was from 2002. It's quite an interesting dataset and it leads to further exploration from the simple suicide/development index question. However, it was also in a format rather untenable to Tableau and required a lot of massaging in Excel as well as a little Python script to get the data into Tableau-acceptable format.
For measures of how developed a country was I used the UN Development Program's human development index. The table contained HDI data from 1975 to 2005, every five years. To simulate 2002 data I linearly interpolated between the 2000 and 2005 values, or just used the 2005 values if data from previous years was not available. The UNDP defines HDI values > 0.8 as "High human development", HDI values between 0.5 and 0.8 as "Medium human development", and HDI values under 0.5 as "Low human development".
For the final table I merely added a column into the WHO dataset containing the HDI for each available country. Some countries did not have an HDI; I set these values as 0 and omitted them from the analysis. It should also be noted that the WHO dataset has measures of "evidence", where more evidence indicates a higher confidence in the recorded data. In certain cases data had to be inferred or modeled.
Correlation between suicide and HDI
My first attempt seemed shocking at first:
Why were there such pronounced spikes? The highest point represented China and the second India; it then became immediately apparent that I was in fact displaying the raw number of suicides versus HDI. The numbers on the left represent total suicides in thousands. Plotting suicide rates against HDI gives a more reasonable chart, where I'm simply dividing the number of suicides by the total population:
Here a correlation between HDI and suicide appears likely, albeit slight. Interestingly, it seems as though most countries have a low suicide rate varying from 0 to 0.01% of the population, yet we start to see some striking exceptions as HDI increases. These exceptions are highlighted in red.
The countries with high suicide rates were rather surprising. They were, in order:
Lithuania, Russian Federation, Belarus, Kazakhstan, Ukraine, Sri Lanka, Latvia, Slovenia, Estonia, Hungary, Japan, Finland
Let's look at this on a world map, where the size of the circle is proportional to the suicide rate:
Note the large circle for Japan, as expected. But look! Eastern European nations and former USSR republics stand out on the map as countries with high suicide rates, although the graphic is a bit misleading because of the density of countries in Europe. Why is this?
Eastern European countries and island nations
I was wondering if perhaps the suicide rate was artificially higher in Eastern European countries due to a greater overall number of deaths. To check this I decided to investigate the percentage of all deaths that were suicides. The results were quite surprising:
The highlighted points correspond to the countries:
Seychelles, Saint Vincent and the Grenadines, Saint Lucia, Suriname, Iceland, Samoa, Sao Tome and Principe, Luxembourg, Vanuatu, Qatar, Maldives, Guyana, Brunei Darussalam, Belize
What immediately struck me was the prevalence of island nations. Here's the percentage of suicides plotted on a world map, again with circle sizes proportional to the percentage:
Indeed, the largest circles are on island nations, with a few exceptions.
Post-mortem
From the analysis, it seems like there is a slight correlation between HDI and suicide. More surprising, however, was the discovery of a high suicide rate in Eastern European countries, and the relatively large percentage of deaths that are suicides in island nations. Tableau made it easy to discover trends and answer questions quickly, and really emphasized the power of data exploration.
