If they're the statistical methods and metrics hold up, or they don't. Also, if you don't want to read my opinion on things, then just grab the GitHub repo and run the end-to-end replication and look at the output data yourself.