Saturday, June 25, 2016

Mysteries of data display


Great health achievement.

But as always I am fascinated by how we communicate things, particularly in numbers.

One of the standing rules of thumb is to watch the axis. Things get hidden by monkeying around by truncating ranges and other such tricks. Also pay attention to legends.

This is a great example. I would have sped right by without those two heuristics. And I don't think there is anything maliciously misleading going on. But the way the data is presented ends up being misleading even though it is a remarkable and positive story.

The first thing that caught my attention was that the two, nearly identical lines are not labelled. Going to the source data, the solid line is for men and the dotted line is for women. A glancing look makes you think that the cancer rates for women and men have plunged in tandem. And, in a fashion, they have.

But having sorted out the line labelling, you then notice that there are two Y axes with different ranges. The one on the left ranges from 1 to 9 odd and the one on the right ranges from 2 to 17. It takes a second to grasp what is going on.

The Y axis on the left is for women and on the right is for men. Let's follow this through. in 1950, women suffered roughly 9 stomach cancers per 100,000 women whereas men suffered stomach cancers at nearly twice that rate at 17 for 100,000 men.

By using different ranges, the chart hides an important fact. Yes, stomach cancers have decline at the same rate but they started from dramatically different points. Men suffer stomach cancer at nearly twice the rate as women.

And what about today? Looking at the chart, you'd think that men and women have converged to the same cancer rate. But they haven't. Men still suffer stomach cancer at nearly twice the rate as women, but both suffer at a far lower level than in the past. Women 1.45 per 100,000 and Men at 2.68 per 100,000. The ration of male suffers to female suffers has barely budged in 50 years.

This is virtually entirely a good news story, the decline in stomach cancer rates. But there is also a communications story in here.

Why use different Y axes ranges? By doing so you hide that men suffer stomach cancer at twice the rate of women and that hasn't changed in fifty years? By charting the data as they have elected to do, they hide a pretty interesting question. Why do men suffer stomach cancer at twice the rate? Is it simply a gender difference? Do men have sufficiently different diets (eat more red meat, drink more beer?) that that might cause the differential. Is the fact that women live longer than men in some way responsible for the differential? Are there other common organ cancers where there is such a strong variance between the sexes and do others penalize women over men, and if so, which and why?

The only reason I can think of to display the data in the fashion it has been displayed is if one were advocating for the distribution of research money. In other words, men would favor larger investments in stomach cancer research than women simply because of the differential.

But the originating site, HumanProgress, doesn't, at a cursory view, appear to me to be an advocacy group.

So a mystery and a good example of the importance of clear thinking about information display. Properly displayed, the data prompts some interesting and unresolved questions. Displayed as it is, the data suggests that the problem is all but solved.

No comments:

Post a Comment