The COVID infection survey is a weekly publication of the six previous week's worth of observations; with model and accuracy updates of those data retrospectively made (i.e. a regions' rate is 5.1% on week n, but on week n-q it was 5.0%). Once an observation falls out of the republication it will no longer be updated.
This model is currently addressed by accreting all editions in a megacube (@rossbowen's name not mine) with an additional dimension to provide publication date. This way a user can select the most recent observation for a given dimension slice by selecting the dimension slice plus the max(publication date). It works, but yuck.
Maybe a qb:Dataset can be described such that weekly editions of the dataset are separate, but the dataset series has a latest which combines the latest available slices so it appears to users that the data is no longer rolling 6 weeks, but the most recent revision of all data is provided in a single cube but no replaced data is.
The COVID infection survey is a weekly publication of the six previous week's worth of observations; with model and accuracy updates of those data retrospectively made (i.e. a regions' rate is 5.1% on week n, but on week n-q it was 5.0%). Once an observation falls out of the republication it will no longer be updated.
This model is currently addressed by accreting all editions in a
megacube(@rossbowen's name not mine) with an additional dimension to provide publication date. This way a user can select the most recent observation for a given dimension slice by selecting the dimension slice plus themax(publication date). It works, but yuck.Maybe a
qb:Datasetcan be described such that weekly editions of the dataset are separate, but the dataset series has a latest which combines the latest available slices so it appears to users that the data is no longer rolling 6 weeks, but the most recent revision of all data is provided in a single cube but no replaced data is.