Tuesday, May 4, 2010

Things that go bump in the stats

Two things have crossed my radar in the past 24 hours that have raised the question about statistical error and, more particularly, when one should act on research findings.

The first was on the ABC's excellent Q&A program where, as usual when politicians are exposed to public questioning, the conversation turned to climate change. Federal Environment Minister, Penny Wong, said that the research indicated that there was a 90% chance that human activity was contributing to climate change (note: global warming has disappeared from the vernacular!).

Segue to a newsletter I received this morning from respected Melbourne research company, Forethought Research, which discussed statistical error in research and what it means in terms of management decision making. There was stuff about p-value etc, which essentially measures the likely accuracy of statistical information.

The point in the newsletter was that companies are generally accepting of 90% accuracy, rather than the 95% that the statisticians would like us all to shoot for. Why is this?

I suspect it is because research is often used to test out management's instinctive or anecdotal suspicions that average performance according to a set of criteria has varied. In my view, a company with its finger on the pulse of what customers are thinking should not be totally surprised by research findings.

Which brings me back to Penny Wong and climate change. This is an area where the variables are so broad, and inputs so vague that anything close to 90% accuracy is miraculous. She put the argument that it would be irresponsible not to take action given the level of certainty around the research.

Penny pointed out that the last decade was the warmest since weather records began. I wonder if this is the 'managerial' anecdotal evidence that enables action on the 90% certainty? There have been warm periods previously. Antarctica used to be covered in forest - but perhaps that was when the Earth was still busy cooling the primordial swamps. The only climate records from then are layers of rock and fossils.

The only certainties I can see in all this is that global resources, though vast, are finite and economists, bless their hearts, still consider that the only valid success measure is growth in GDP. We talk about sustainability and still breed like rabbits.

More significantly, western democracies seem to have rendered themselves virtually impotent to do anything about it. Opposition politicians drool at the chance to harpoon governments if GDP declines, interest rates increase, new home starts fall and so on. Sometimes these are necessary evils. It's why the Chinese may be the best-positioned to act on climate change and make a difference. The are big emitters and special interest lobbies and the loonie fringe find it harder to get a foothold there.

In the end, climate change cycles outlast political cycles and will ultimately prevail. And you can bet on that with a high degree of accuracy.

No comments:

Post a Comment

Any thoughts on this, drop them here...