That tension when your boss writes you "I've found a problem with your method, we need to discuss this soon" without stating what's wrong. [audible sweating]
I once read a paper in the applied political science literature that showed that just adding more variables to a multiple regression model did not necessarily get you 'closer' to causality.
I can't seem to find it now, and I don't remember the author's name, the title, or the journal. I do remember that it was applied methodological advice for political scientists. I'm pretty sure it would be >2000; it wasn't that old.
@gung Applied political science ... maybe someone like Jackman or even Gelman? It sounds like the sort of thing Gelman could say. I'd be interested to see it if it turns up.
I am a bit surprised to see that this bountied Q stats.stackexchange.com/questions/315979 does not have a decent answer. Looks like a pretty straightforward bur reasonable question of general interest. @Glen_b maybe?
@amoeba Yeah, I saw that when it first posted. I could try to post an answer when my present period of heavy-workload is out of the way - a few more days at most, I hope - but I'd have thought gung (or any of a number of other highly accomplished answerers) would be better at that one than me.
The bounty is not of interest to me -- in recent years I actually avoid bountied questions, hoping that others will be moved to post a good answer by the bounties and get the reputation; generally I don't want to get in the way of that positive effect of bounties by answering a question under bounty and risking getting the bounty over someone who might have posted at least as good an answer as I would if I had not.
I do sometimes answer after the period though if I think there's anything useful to add.