Researchers propose using Bayesian statistics to fix the 'replication crisis' where past studies fail to produce the same results when repeated.