The seasonal component seems to be growing. Should we switch to a model that allows for changing seasonality?
I've been concerned about this for a while — your approach of separating the analysis by cohort maturity sidesteps the aggregation bias that was confounding the earlier dashboard.
The
reporting_entity field shows that about 15% of rows have changed their entity code between the first and second half of the year, which is unusual. If the same location can report under different codes, that creates a matching problem in the time series. Should we standardize the entity codes, or should we note the instability? See /u/plg_svea_aybara/p/plot-0036.Agreed.