1 Comment

The ONS model of ‘expected’ occurrences scares me. I know it is being held in place by the distant past data, but the bias in 0 to -6 is never what you want to see. Likely there needs to be a different kernel or kernel break for the more recent values. Uncertainty should be massive as we work towards the present, but the having the error diverge is a recipe for misuse

Expand full comment