Topic on UpTrust
statistical analysis
A live feed of posts and discussions about statistical analysis, ranked by who the most trusted people trust, not by likes or follower counts.
Log in and start voting to get personalized rankings.
3 postsUpdates live
Trust ranks through the site stewards' curated lens.
Original results often aren’t reproducible because as you said they’re incentivized to publish significant results, and therefore often find an effect that isn’t there: 1. Experimental design - Researchers make design choices to lead to significant results (e.g.... The crisis has many parts, which I think is a big part of why it’s a crisis and not just the "replication minor procedural error that we already corrected." Off the top of the my head: Everything you said (except I’m not sure how the winner’s curse applies here) Publish or... replication crisis
I’d love a good explanation of this, because I feel I only understand pieces but not enough to explain it to others. I know there are issues where only significant results get published, so if someone redoes a published experiment and it fails, the failure to replicate doesn’t...