Correlation doesn't imply anything about causation, though the article itself is all about how maybe full-fat dairy causes lower obesity.
My point is that perhaps obesity causes less consumption of full-fat dairy (because "I'm getting kind of overweight; I'll get the low-fat one") OR there may be another factor that's causing both effects -- e.g., organic yoghurt is much more likely to be "natural" (full-fat), and people who eat a lot of organic food as part of choosing their diet tend to eat heathier (less fattening) meals.
I'm making both of these up -- but both situations explain a correlation/association between full-fat dairy and lower obesity without full-fat dairy causing lower obesity. A good study needs to rule this kind of stuff out to be useful.
Often, unfortunately, the study itself mentions various confounding factors, alternate explanations, etc., and the reporting just ignores all that, cherry-picks the desired "result" and poses the headline as a question to be safe.
If they're tracking two measurements, and observing a correlation, "people who get obese stop eating full fat yoghurt" needs to be ruled out of the effect, somehow.
But why don't you just go find an explicit flaw in their methodology, rather than assuming there must be an obvious one there?
From experience, I tend to suspect the article is at fault, not the study, though the error of "spinning a study to make it appear to prove something it didn't" is unfortunately rather more common than something like "didn't mention anything about how the study showed causation".