↻
Claas Voelcker reposted
Berna Devezer
@devezer.bsky.social
📣 The difference between replicable and not replicable is not itself scientifically replicable. 📣 New work with Erkan Buzbas, showing that verdicts such as "X% of results replicated" are based on an inferential machinery that doesn't work. arxiv.org/abs/2604.26268
AI Weekly's analysis
→
- Under a benchmark model, small variability in replicability rates creates an irreducible variance floor that no amount of replication can eliminate.
- Under an operational model, one binary verdict per experiment carries no information about between-experiment heterogeneity.
- The authors reanalyse Many Labs 4 and conclude the replication crisis, if there is one, cannot be established by the methods used to declare it.
Read full analysis →
View on Bluesky →