As you may have noticed, I am a big fan of visual data analysis for its speed, clarity and for the way it promotes data exploration.
Last time, I was talking about dynamic heat maps and how they could analyze and present, with clarity, large amounts of data with multi-dimensional characteristics.
This time I’d like to talk about “table lenses”. The graphic presented below is an example.
This is a set of bar graphs on birth rates, population ages and health expenditures for 208 countries. In this example, we might be wondering if the World Bank’s statistics show a relationship between birth rates, population aging and health expenditures.
What the graphs present is, from left to right is:
infant mortality rate
% of population aged 0-14
% of population aged 15-64
% of population aged 65+
health expenditures as % of GDP
health expenditures per capita
The graphs are all sorted on the first graph, i.e. by birth rate. Doing this allows us to “eyeball” the correlation of these data series. For extra emphasis on the analytical issue in question, the bars are coloured according to each country’s health expenditures per capita. This is a lot of information conveyed quickly, clearly, and dramatically.
Clearly, birth rates are correlated with infant mortality rates (first and second columns to the left). Although infant mortality rates are higher, generally, in high birth rate countries, the age distribution in those states is still concentrated in the youngest cohort. Countries with lower birth rates generally have an older population (see the middle three columns). A reasonable inference, requiring, of course, further investigation for confirmation, would be that the lower birth rate countries, while not adding as rapidly to their younger population, also support longer life spans.
Are these population characteristics correlated with health expenditures? Interestingly, they do not seem to be correlated with health expenditures as a percentage of GDP (second column from the right). They do, however, appear to be correlated with health expenditures per capita (last column to the right). The latter result is as you would probably expect, and it is reasonably clear here, even if the correlation is not anywhere near to being perfect.
The lack of correlation of health expenditures as a percentage of GDP and expenditures per capita is also of interest. Evidently, even if a country were to spend much of its wealth on health, because of low GDP and/or a large population it may still not be able to spend enough per capita to achieve better results in infant mortality and life expectancy.
Of course, it is worth checking this statistically, but there is little doubt that there are significant correlative relationships in this data, as revealed quickly and clearly by the table lens.