From a statistical viewpoint, what are the dangers with treating a negative variance from the EMS method as zero? And why would you use REML or Bayesian Instead?
For EMS vs REML, take a look here, particularly the section ''REML vs. EMS . . . '. For the Bayesian approach used in JMP, see Laura's 2015 Discovery paper.
There are no labels assigned to this post.