How is it possible that the higher levels, have more unique series than the base level? How does it aggregate in that case?

copy paste from chatgppt:

In a "clean" hierarchical‐forecasting setup (i.e., where each parent node is strictly the aggregate of its children), you will never see more unique series at a higher level than at a lower level—because by definition the higher‐level series are sums (or some other aggregation) of the lower‐level ones. In other words, each leaf series belongs to exactly one child‐group, each child‐group rolls up into exactly one parent, and so on. Since you can only aggregate "up," the count of distinct series strictly decreases (or at best stays the same) as you move from bottom (leaf) → top (root).

No aggregate level can have more series than it's lower level children. Of course if you define many levels with a non-strict hierarchy, the total number of series in aggregations may exceed the total number of base series.

ah, interesting. i need to think about this! thx