,
4 tweets,
1 min read
Read on Twitter
One other point on this paper. Something I still struggle with: What happens when the stats that are best for a paper are stats that 95% of the field and esp readers won’t understand? We tried out best to explain them, but if we’re honest, even all of us don’t fully understand.
Getting an editor + reviewers who were savvy was key. But it is unrealistic that everyone will understand increasingly complex analyses & it’s unlikely that non-scientists will. What do we do? Is it ok that so many people do not understand many details of the results section?
Simple stats (t-tests, regressions) told the same story as the fancy Bayesian analyses (though the latter let us support the null). Part of me wondered about putting both the wrong but simple ones and the right but hard ones so more people understood.
Instead we put the right stats and just assumed lost readers would move to the discussion. But this doesn’t seem ideal. Have people grappled with this? What’s a solution that is realistic?
