This is an impressively bad photo haha, but the work seems really interesting! Any chance of getting access to the poster somehow?

So, check it out if this seems useful to you! Done with @xallysie.bsky.social , @eugenekofosu.bsky.social and Gabe Nespoli
/end

That said I've since learned that we are basically brute forcing a math proof of the law of large numbers and approaching population parameters. Convinced this is true, but for me the tool far easier to use practically and we still use it internally all the time. Other approaches in the paper
5/n

Initially an internal tool we made a good while ago to answer some qs and guide data collection. But we made it public with a little write up. And then that white paper got more citations than much of my real empirical work. So seemed useful to folks, and we decided to write it up.
4/n

Having stable ratings trims out some error from your overall model, optimizes data collection so you don't waste money, and helps you plan how many ratings to collect for x rating (for example)
3/n

For people who collect a bunch of ratings, combine them in some way (avg), and then analyze those averages. These avg ratings are "stable" when more ratings wouldn't meaningfully change the avg. When more raters agree on a rating, stability is achieved faster, but this varies by stim and trait.
2/n

mmmmm good figure