Alcohol and breast cancer

[pioneer]

I heard about this on the TV and thought it was a good example of good statistical studies leading to irrational results that will be overturned with the next set of studies. The reason I say that, statistical studies are based on a blackbox, where we measure input and output and then create a correlation. But not knowing what is in the blackbox typically leads to short term hype followed by another study that will say the opposite.

 

When I thought about this, I came up with the perfect analogy to describe how something mathematically valid can still lead to the wrong conclusions. For the blackbox in this study, we are going to use a soda machine. Our inputs are coins or paper money, and our output are cans of soda. Since it is a blackbox, we don't need to know what is going on inside to be able to run this statistical experiment.

 

Based on this scientific study, coins are 50% more likely to produce a soda can on the first try. This is a blackbox study that needs to stay closed. If we had opened it instead, we would have seen the paper money scanner needs the money in 1 of 4 orientations, no folds and not to wrinkled. Based on our blackbox data, one possible theory for why coins output more soda cans on the first try is, metal-attracts metal, which is why coins output soda "cans", with a higher level of reliability.

 

Based on the scientific data and the black box, that is not a bad theory. If we could look inside the blackbox, one could refute this, but the blackbox rules in such studies. The next study, decides to strengthen this theory by assuming since metal helps metal, than maybe paper money does better with plastic soda bottles, since they are both based on carbon. After that set of experiments, the opposite occurs. The coins are actually 51% better for plastic??? This study concludes, it is not "like attracting like", but an oil and water affect, where the metal repels plastic so they come out better. The 1% is within the margin of error so we keep both theories.

 

In real science, the rational theory comes first, then you run the experients to prove your line of reasoning. Nobody would have come up with metal coins attracting metal soda cans and then to try to run that experient. But with blackbox studies, one can come up with anything since one only correlates input-output data, and reasons later. Relative to the alcohol study, the next will show alchohol repels plastic soda bottles.

[SkepticLance]

Actually, there is nothing new about the alcohol/breast cancer link. This has been shown before, and the latest study is just confirmation.

 

The results can, of course, be misused. Let me put it into perspective.

 

Define a standard alcoholic drink as a large glass of wine, or the equivalent in other alcoholic drink. Consumed on a daily basis, each large glass increases risk by just over 2%. Among women, the lifetime risk of breast cancer is 9% on average. Thus, if a woman drinks a full bottle of wine per day (4 glasses) or the equivalent, she doubles her risk. That is, her lifetime risk becomes 18% or less than 1 in 5. Since most breast cancer these days is non lethal, the consumption of alcohol is not quite the death sentence that some people will make it out to be.