New paper estimates the replicability of >14,000 psychology papers using machine learning. They found:
1. "Experiments replicate at a significantly lower rate than do non-experimental studies"
2. "authors’ cumulative publication # and citation impact are positively related to the likelihood of replication"
3. "media attention is positively related to the likelihood of replication failure"
via https://www.pnas.org/doi/10.1073/pnas.2208863120
#Science @academicchatter #Psychology #MetaScience #ReplicationCrisis #SciComm
Follow
@erinnacland Didn't read the article. What do they mean by replication of non experimental results? What exactly are they replicating?
@erinnacland Ah, alright! Thanks for the explanation!
@rastinza Non-expirmental research includes studies where researchers aren't in control of the treatment or independent variable(s). For example, if look at how income is linked to later increases in criminal behavior, but aren't manipulating how much money the participants earn, that is an observational study, i.e., non-experimental design.