Mon: DG XXXVII: Lumosity fined $2 million for deceiving customers about its “brain training” programs
Tues: “if you add a few more variables, you can do a better job at predictions”
Wed: Stochastic natural-gradient EP
Thurs: A new idea for a science core course based entirely on computer simulation
Fri: Oooh, it burns me up
Sat: What is the “true prior distribution”? A hard-nosed answer.
Sun: Risk aversion is a two-way street
I don’t know where to put this comment or if you usually have an open thread or what, so I’m just putting it here.
Has anyone else noticed that it’s becoming common for papers to report an effect size, then run some controls, and say “a statistically significant effect remained post controls” without mentioning whether the size of the effect changed? I’m pretty sure this is becoming a hot new way for people to lie without lying, in a way that’s difficult to call them on. Makes me mad.