Correlation vs causation

Today's Boston Globe commits one of the classic blunders of inferential statistics -- confusing correlation and causation. The Globe had the following headline:

Beverage reseach tied to corporate dollars

Conflict of interest seen when industry finances studies

The article describes a study from Childrens' Hospital that finds that nutrition research sponsored by the beverage industry is more likely to yield results that favor the industry than are studies that are not funded by the industry (64% versus 46%). The article concludes that industry money biases nutrition research.

Well, it may very well be true that industry money biases nutrition research, and I have no trouble believing that. Unfortunately, this data doesn't prove it, and indeed, could just as easily prove the opposite. Instead of industry money biasing research, it's plausible that causation really works the other way around -- nutrition research may bias the allocation of industry money.

How would this work? Instead of paying off researchers, it may be that dollars get allocated in a more benign way: corporate sponsors scan the research horizon and place their dollars behind scientists and research approaches that are already favorable to industry positions, and thus, are more likely to continue to generate findings that support industry positions.

Industry may not be paying off researchers so much as placing bets on which research is going to go their way. And the problem is, the data would look the same either way -- funding would be correlated with results under either scenario.

This same problem of trying to separate causation from correlation also exists in analyses of corporate contributions to Members of Congress. While the presumption is that oil industry money influences Congressional voting on environmental laws, for example, it may also be the case that oil industry money really just rewards Members who would vote against environmental interests anyway.

I think that researchers would be wise to somehow separate themselves from this money, because it hurts their credibility even if it doesn't affect their research, it's becoming increasingly transparent to the public, and the NIH extramural research model relies on their integrity.

I would also like to see Members of Congress do the same. Unfortunately, credibility may not matter quite so much for them.......