Now it tests both the uncertainty in the mean and the uncertainty in
the observation. Previously, it only tested the observational uncertainty (think prediction interval), which is a simpler calculation. That coverage gap could have allowed a bug in the more complicated mean uncertainty calculation.
Now it tests both the uncertainty in the mean and the uncertainty in the observation. Previously, it only tested the observational uncertainty (think prediction interval), which is a simpler calculation. That coverage gap could have allowed a bug in the more complicated mean uncertainty calculation.