This is a nice example of a statistically significant effect that has negligible predictive value, but that is sold precisely on that value. The title of the pa

This is a nice example of a statistically significant effect that has negligible predictive value, but that is sold precisely on that value. The title of the paper--"Why Susie Sells Sea Shells at the Sea Shore"--is clearly intended to suggest that implicit egotism results not just in a statistically significant effect, but a usefully predictive one. But the questions of where Susie lives and what she does for a living are not answered in any ordinarily meaningful way by the effect the paper identifies. If I have a coin with a few percent bias toward heads, it would be easy to prove this in a couple of hundred flips, but if I were to say every single time it landed heads, "That's because it's not a fair coin" people would look at me funny. And yet we do this all the time with far weaker causal factors: every significant weather event in the past decade as been reported as "due to global warming", even though we can be pretty sure that even without global warming there would still be weather, including extreme weather, and even people who take the detailed results of climate models more seriously than I do don't claim they predict shifts in extreme weather event distributions that are detectable on a decadal timescale. Moral: we should be cautious about jumping from a statistically significant effect on distributions to a usefully causal account about individuals, even when it is pretty likely that correlation does imply causation, as is the case here.