Death of Trees Correlated With Human Cardiovascular & Respiratory Disease

eldavojohn writes "PBS's NewsHour interviewed Geoffrey Donovan on his recent research published in the American Journal of Preventative Medicine that noted a correlation between trees (at least the 22 North American ash varieties) and human health: 'Well my basic hypothesis was that trees improve people's health. And if that's true, then killing 100 million of them in 10 years should have an effect. So if we take away these 100 million trees, does the health of humans suffer? We found that it does.' The basis of this research is Agrilus planipennis, the emerald ash borer, has systematically destroyed 100 million trees in the eastern half of the United States since 2002. After accounting for all variables, the research found that an additional 15,000 people died from cardiovascular disease and 6,000 more from lower respiratory disease in the 15 states infected with the bug, compared with uninfected areas of the country. While the exact cause and effect remains unknown, this research appears to be reinforcing data for people who regularly enjoy forest bathing as well as providing evidence that the natural environment provides major public health benefits."

Comments

Before assuming "they didn't control for"

[pepty]

and posting your indignant observation, please check and see if they did.

Re:Before assuming "they didn't control for"

[nadabu]

Sorry but the very phrase "After accounting for all variables" when doing statistical analysis on any complex real-life scenario is laughable. We don't even know all the variables, much less have rigorous data for them all.

I think their theory is probably right. It makes a lot of sense and the data we do have does fit. But this is statistics, not science; correlation, not proof of causation. It is far from being without value, but it is also far from being conclusive or thorough. It is merely as thorough as it could be given available data.

Re:Before assuming "they didn't control for"

[gstoddart]

There's things you think you know that are obvious, and then there's actual science.

Sometimes, you need the proper study just to verify your hunch isn't entirely wrong -- everything else is an anecdote or a guess.

Re:Before assuming "they didn't control for"

[Jane Q. Public]

"Want to bet that this will be exhibited soon as a poster child of spurious significance and poor statistical analysis?"

As Darrell Huff, author of the 1954 classic How To Lie With Statistics pointed out, the salaries of Protestant ministers at the time was very strongly correlated with the price of Jamaican rum.

The point being: so what? A correlation is all well and good, but the chances are overwhelming that it means exactly shit.