With large enough LLMs you don't need randomisation

Show Post

"250,000 respondents can produce an estimate of the population mean that is no more accurate than an estimate from a simple random sample of size 10" reminds of recent Nature journal paper with N=billions but lacking the studied variable so the study is actually nonsensical doi.org/10.1525/coll...

Show Post

What’s the date that people celebrate the anniversary of the tweet about “with large enough samples you don’t need randomisation”?

Show Post

There's some important work on surveys that shows, like, a well-collected N=1k sample can be much better than an N=100k convenience sample and I guess this hasn't been picked up in metascience or psych yet

Show Post

Aw, thanks Ioana. It was just as nervewracking as it was wonderful, but with every day that passes, this asymptotes to "just wonderful".

Show Post