Hacker News Reader

Comment by scott_s | Hacker News Reader

This is a terrible graph. It took me five minutes to figure out how everything is related. Data such as this is much more naturally plotted as an x-y scatter-plot. Thickness of the datapoints could indicate number of doctor visits. Instead, the graph eliminates the x-axis, but retains the space. Very strange choice.

View on HN

Agreed. Here's the same data in a scatter plot (second chart down). Far, far easier to grasp.

http://www.stat.columbia.edu/~cook/movabletype/archives/2009...

Thank you. As I suspected, I can immediately read the scatter plot. And the correlation that I wondered about is indeed there.

The problem with the scatter plot is that you lose the average cost/average life expectancy info. On the original plot, the UK corresponds roughly to average health care costs and average LE. The slope shows if the country is getting value for money. The upward slope of New Zealand shows that they are spending less money per person than the average yet getting better than average results. The downward slope of Denmark shows that they spend more than the average for inferior results. That information is lost in the scatter plot.

But it could easily be added in by having horizontal and vertical lines added that read "average healthcare cost" and "average life expectancy."

IMO, the slope of the lines in National Geographic chart show the Spending/LE relation better.

I agree that it's unfamiliar, but once you've figured out how to read it, it strikingly demonstrates the degree to which the U.S. is an outlier.

My claim wasn't that it's unfamiliar - which it is - but that's it's terrible. What's striking is the data, not the visualization applied to that data. I think the effect would be just as dramatic with an x-y scatter plot, and it would have the benefit of being easier to read.

This graph also eliminates seeing if there is a correlation between money spent per individual and life expectancy. On a scatter plot, if there was a correlation, we would expect to see the data fall in some line. If a data point was outside of that line, then it would be obvious.

Actually, this seems like a terrible visualization because it's taken out of context, and was originally presented in a magazine. You can see where the article text would go, the United States label is hidden away near the binding and far in the upper margin, and due to expected info location/"eye drift" one doesn't find the U.S. at first.

Wish I could provide a scan of the page as it went to print, but I only saw it while visiting family. Regardless, the presentation is clearly to cause that double take rather than just to efficiently convey the information.

the graph gives an important ratio that's hard to capture in scatter plot

age/cost (the slope)

also, how do you identify your datapoints in scatter-plot neatly?

to me this graph gives surprisingly rich information packed

I don't understand your first point; the relationship between two variables is more clear to me from a scatter plot. And that's something else that bothers me about the graph: when I look at a graph with lines, I expect to see trends. But the lines in this graph are actually single data points.

We're on the web, so it would be neat to be able to use the mouse to hover over data points to see labels.

Scatterplots are usually necessary to check correlations. See

http://en.wikipedia.org/wiki/Correlation_and_dependence

for a famous set of example scatterplots first designed by a statistician warning students of commonplace errors in interpreting correlation coefficients.

in the scatter plot this information is more clear to me, also - in the scater plot if I draw minimal squares line, I can immediately see the countries that are "better thhan average" and "worse than average" with their efficiency of healthcare.

Reply on news.ycombinator.com