Dates in time series

[jeanpaulrsoucy]

Generally, time series only go up to the date the data in a particular PT was actually updated. However, there are exceptions: datasets updated using active_cumul and datasets pulled from the covid19tracker dataset, since these get values for every day.

There are two ways of potentially addressing this:

• Hard code the day(s) of the week that data get updated
  • ...but sometimes updated occur on different days due to holidays
• Automatically remove data up to the most recent date the data changed
  • ...but sometimes the data are updated but legitimately don't change (e.g., there were no deaths the past 7 days)

A solution might be to combine the two approaches: keep data up to the hard-coded day of the week or the most recent date the data changed, whichever is later. This could be incorporated into the postprocessing step (#1).

[jeanpaulrsoucy]

We should also look into truncating vaccine coverage (#21) and vaccine administration (#47) time series (i.e., beginning from a minimum date rather than all time series beginning at the start of the vaccination campaign.

[jeanpaulrsoucy]

This is no longer an issue with active_cumul data and should no longer be an issue with covid19tracker data once these datasets are replaced for current AB, BC, and PE hospitalizations/ICU datasets (#82).

[jeanpaulrsoucy]

Now, the only hosp/ICU dataset connected to covid19tracker for current updates is PE. However, covid19tracker has apparently ceased updates for this province's dataset, so this issue may be considered closed.