[1801.05042] Centralized "big science" communities more likely generate non-replicable results
Very interesting preprint suggesting that "Claims robust to multiple, socially independent investigations have a 55% higher probability to replicate than those studied by small clusters of overlapping collaborators." by Danchev, Rzhetsky, & Evans
Centralized "big science" communities more likely to generate non-replicable results
Papers by James Evans et al. on why centralized "big science" communities are more likely to generate non-replicable results & the one on the differences between small and large teams
3. "Centralized "big science" communities more likely generate non-replicable results" by
. This massive undertaking compared literature claims to
L1000 results to find associations with the probability of replication.
Decentralized research communities involve more independent teams and use more diverse methodologies, generating the most robust, replicable results.
What do you think of this
It impacts #OpenScience to some extend as sharing uniform methods and materials across a community could lead to built-in fragility -> results less likely to be robust than in highly diverse/disconnected settings