Hearing "Throwing individual level data away reduces the amount of information" described as a fallacy worries me a bit. For one thing this statement is true (for any technical definition of "information), but more importantly I feel it's ignoring lots of progress in getting people to model individual-level data in hierarchical models, rather than just e.g. a regression on group means. Throwing away data might be worthwhile for any number of reasons (e.g. computational or to make a plot more readable), but it *is* throwing away information. And if you can, it's better to build a model at the level of individual data than aggregates. I wasn't at the talk and don't know what was said, so maybe all of this was made clear, and the "fallacy" language is just supposed to be a good sound bite.