Rejection of science squared

Or: Computer modeling is not science

Fat people are now being blamed for contributing to climate change by breathing. Actually, it’s worse than that. Headlines are saying: “Fatties cause global warming.”

Hundreds of news stories this week are based on the latest carbon dioxide paper just published in the online version of the International Journal of Epidemiology. Its authors were Phil Edwards, senior lecturer and lead organizer of the module “Statistics for Epidemiology and Population Health,” and Ian Roberts, both in the Department of Epidemiology and Population Health at London School of Hygiene & Tropical Medicine. [There’s that new field of population health again.]

As JFS readers will remember, professor Roberts had written the article in New Scientist, published in 2007 also around Earth Day, titled: “How the obesity epidemic is aggravating global warming.” It had accused fat people of killing polar bears and had begun by stating: “WE KNOW the world is warming and we know humans are causing it...”

While being called a study by epidemiologists and the media, reading the actual text of their new paper reveals the methodology:

“The upward shift in the population distribution of BMI could also have important environmental consequences… we assume… since it can be assumed… the equations used were… to estimate, we assumed… estimates were applied… these broad assumptions which affect predicted energy expenditure… since it can be assumed… on the basis of these assumptions, we estimate that… might therefore be responsible… we would expect the overweight population to have higher transportation fuel energy use… to estimate, we assumed that… we assumed… our model estimates that… expresses the idea that… would not be surprising…might reasonably expect…to estimate, we have assumed… if we assume… our estimates… it is likely… might therefore be considered… if we assume… if… it is likely that… We used a normal (Gaussian) distribution to model... and a log-normal distribution to model the skewed distribution reflecting a higher prevalence of obesity and morbid obesity in the ‘overweight’ population. These are theoretical statistical probability distributions, which may not be expected to describe perfectly the shapes of the population distributions of BMI observed in high-income countries… In our model, we have assumed... If...then we would have over-estimated average energy expenditure in the overweight population.… When we assume plausible…

We have estimated the additional green house gas emissions due to increases in population adiposity. In doing so, we have made a number of assumptions all of which can be questioned.”