Slices of higher-dimensional Counts and Probabilities should yield new Counts and Probabilities

[kahaaga]

Relevant outcomes also must be propagated

[kahaaga]

On second thought, this is speculative.

Slicing a Probabilities array means the Probabilities will be re-normalized. In most use cases, that is probably not what a user would want.

The reason I thought of it is that viewing a probabilities array should look nice (outcomes should be displayed in the margins) even when slicing. But I don't immediately see how this can be achieved without breaking the contract that the probabilities must sum to 1.

[Datseris]

Does this has any use to begin with...? Marginal propabilities are not slices anyways.

[kahaaga]

For counts it has meaning to slice. For probabilities less so - it would have to be a new type.

In terms of pretty printing, it does have meaning to slice, because one may want to look at some subset of the data other than the default printed slice. But I don't consider this important enough to postpone 3.0. It can be done at a later stage, perhaps by just creating a view that explicitly lets the user know that they are looking at a subset. If we deem it necessary.