[QUOTE=TimeWinder]
My initial impression on reading the news release was “huh - scientists have discovered that people more likely to gain weight are more likely to drink diet pop. How…incredibly obvious.” Sure, there are people who drink diet pop for the taste, but I bet that population is dwarfed by the ones who drink it for weight control. And in that population, you wouldn’t expect a few percent decrease in calories from a pop change to make all that much difference to their general trend toward increasing weight.
[/QUOTE]
Speaking as someone with some training in human behavior stats, this sounds extremely possible. Two points:
–correlation does not equal causality. In other words, a relationship does not mean one thing causes the other. It could be chance, or, more often, what is probably true in this case, that these are both effects of the same cause: dieting means drinking diet drinks, but it also means low chance of success.
–when you are talking about food, nothing is truly independent. I doubt there is any food that isn’t somehow connected with ethnicity, nationality, income level, or some other factor that could also affect your general health. It reminds me of some of the salmon-eating studies: that stuff is expensive in some of the places they checked, and more expensive means richer people buy it, which means in general access to better health care. That doesn’t mean salmon aren’t healthy to eat – there is other scientific evidence of the benefits of fish oils --but that the true effect probably isn’t known. There is NO true controlling for other variables is this sort of thing. Most of the time, they control for only linear effects (e.g., effects where you can say for every one unit of this stuff, life expectancy always goes down (or up) by this fixed amount of time). Linear effects can account for some of the relationship, but not all of it. In fact, removing the linear effect can sometimes make an effect stronger. When they do remove some non-linear effects, it is still always something predictably curved, leaving a lot of variance unexplained and uncontrolled.
Moral: ask yourself if the effect could have have some non-food explanation, like TimeWinder did, especially if there isn’t something experimental involved. If they’d taken two sets of overweight people, gave one group water and the other diet pepsi, and found a big difference, start to listen. If they isolate a chemical in diet pepsi clearly causing people to overeat, start to listen. Until then, say, 'huh! I’ll be interested to see more data."