Friday, 13 February, 2009

Dr Rosling's amusing stats

Search on Data Visualisation and you're soon to find Dr Rosling, his amazing stats and his fawning tribe. A little thought shows both the good doctor and his tribe are amusing, but hardly amazing: the former, for what he does; the latter, for swallowing his gobbledygook.

He begins with a test for his students and colleagues. He takes 5 pairs:
1.Sri Lanka/Turkey
2.Poland/South Korea
5.Thailand/South Africa

And asks the test-takers to pick the country with the higher infant mortality from each pair. They score worse than a chimpanzee would (plain chance). He announces that nobody knows nothing.

Now, how fair was that test? Let's look at the actual figures as per the UN:
1.Sri Lanka (11 deaths per thousand births)/Turkey (27.5)
2.Poland (6.7)/South Korea (4.1)
3.Malaysia (8.4)/Russia (16.6)
4.Pakistan (61.5)/Vietnam (19.5)
5.Thailand (10.6)/South Africa (44.8)

The mode, as per the UN data is 13.4, while the median, 21.5. The range is 157.4.

Our pairs are near the middle. Yes, the smaller number is about half the larger one, but that not alone is not enough to take people to task for failing. I mean, it see the students' and other teachers' failure as proof of Western ignorance is far-fetched.

(Can the average student in a US History get these presidents in chronological order: Van Buren, Harrison, Tyler, Fillmore, Pierce, Hayes, Arthur... Probably not. Is the average student a dunce? Probably not.)

But this is only the beginning. What Rosling does next would be laughable had his conclusion not been so dangerous.

He proves that the world has become a more egalitarian - by measuring the incomes on a logarithmic scale!

Now, why would you do such a thing, unless the differences continued to be of orders of magnitude? His very use of the logarithmic scale gives away that reality is the opposite of what he's supposedly proves!

What follows is worse. He shows how developmental variables change with per capita income, the latter still being measured on a logarithmic scale.

Of course, everything improves with rising income! Rosling warns us against generalization, as a good teacher should. But the damage is done. Trade = increased income = wonderful things: babies stop dying; people start to use contraceptive; we live longer; heaven descends on earth.

Or does it? Do the figures really show that? Or do they show that economics probably doesn't matter. Because measured on a sane scale, poor nations stay more or less where they were (and that matters - more in a moment). Yet development measures improve dramatically.

And, in all probability, that had far more to do with science than all the business that ever was. At least, science made the poor man's life far less miserable than his grandfather's.

Meanwhile, income did crawl up. And by using on a logarithmic scale the enrichment looks phenomenal. It's smoke and mirrors! As one commentator wrote at the TED site, "Using such a scale, you can prove anything." (In a log scale, men and elephant are brothers, with the solar system as their first cousin. So what?) The tragedy is that critics of his antics are few and far between; admiration overwhelms.

PS: Can it be thus: There is high correlation, but no casualty; rather both increase in income (meagre) and betterment in development measures (fantastic) come from the same source - science?

No comments: