Conversation
Replying to
Selection bias is fine as long as you understand and communicate clearly what your selection is. This is more or less my whole problem with statistics: People don't use it right and are dishonest about the limits of their data, and then they extrapolate.
4
Replying to
Selection bias being inherent=\=Twitter polls are representative of gen pop or that those results allow for drawing broad conclusions.
1
Replying to
“But it’d be a way better sample if it were bored undergrads low on cash.”
SMH. Idk that I’ve ever found a social study I couldn’t poke a hole in — eventually I had to get over it so I wouldn’t keep rejecting the stuff I intuitively disagreed with.
Replying to
My favorite twitter genre is the criticism of observational studies for selection bias, with RCTs described as the solution.
Replying to
ah ok, I think this makes sense is psychology but much less so in other social sciences. When I use GDP data I'm assuming the selection bias is very small, similarly for other administrative data. But i think my response to this is that isn't econ that's had a replication crisis
Replying to
Selection bias is very much a real thing in many studies, yes. However, you can't just handwave it away like you're doing in this blog post. You need to correct for it to try and remove the bias, because most statistics only works on a representative sample.
1