I'm also prickly about the idea that survey experiments reveal which messages "poll well" in some kind of popularism sense.

They just tell you which message has the highest ATE on vote choice, which may or may not be the messages that people "like" the most

I've summed up what I personally know about attention here. I don't yet think we have the design worked out like we do vote the ATEs on vote choice. But I don't think a better experimental design is out of reach, I just don't think we've coordinated on it yet.

will no one think of the visualizers [smh]

I suppose that first paper might have been the weirdo alpha of the time we make a type I error, but I suspect error rates here are not nominal...

Whoa that VOLCANO! (did you get the underlying data?)

love this idea and I see your design problem re: significance on both sides of zero.

I think conditioning on significance from a one-tailed p-value would really make the point so well!

obligatory C&H

Thank you for sharing -- I was unaware of this framework for choosing among replication targets. I think I stand with the critics. Of course I'm left still not knowing how to choose among empirical estimands!

Separately, this figure is amazing:

Truly! and thats *among* the studies that didn't already post their data.

Easily a majority of the studies in this area were already public.

Journal policy changes, norms of transparency, it's all happening!

no no it's not THAT good :)

first prompt was:

best practices according to hadley wickham for starting a brand new r package devtools roxygen etc.

we want the video

touché :)

Don't most people? I think it's the submission portal clunkiness to blame not the shirking of responsibility. Only time people have turned me down is when they are not going to be at the conference, so my view is, assign the best discussant you can think of for each panel.

surely it's the kurtoses of production and surely its the fat cats who own those fat tails

Ack -- not norms, behaviors!