Smoking and dementia - but is it True?

I am as strong an anti-smoking fundamentalist as they get, but a recent story in the New York Times about a study that found a relationship between smoking and dementia (which I am naturally primed to accept, given my dislike of smoking) caught my attention as a perfect example of how DV409 students can put their analytical tools to work.

There are several things to think about here. First, this is a cross-sectional study on a time-series issue. Are people who smoke now a good proxy for the future of someone who currently doesn't smoke but might take it up? Second (related to the first), the authors included some control variables (the NYT does not list them but provides a link to the actual paper), but are these sufficient to control for all the characteristics that differentiate smokers from non-smokers and that might also be correlated with dementia? Finally, could the relationship be the result of reverse causality? Could people prone to dementia have related neurological issues for which smoking provides some relief, so that smoking is a form of self-medication? This is not necessarily as much of a 'high tide and full moon' (i.e. very low probability) possibility as you might think; there is evidence that many of our behaviours, from smoking to drinking to chocolate-eating (my particular vice!) have at least some element of self-medication to them.

The answers to these questions must be derived from our knowledge of biology and behaviour, but the fact that we pose them in this way at all is a result of our methodological understanding.