Add alternative correspondence methods

[mjkerrison]

My understanding of the get_correspondence_abs function is that the population-weighted correspondence is just from the ABS itself. I'm wondering if there's appetite/scope to include a stack of alternative and/or more updated correspondences:

1. Self-calculated population-weighted correspondences, to provide updated (but cruder) versions while waiting for the ABS to release updated ones based on mesh blocks
   a. Both pro rata (one-to-many)...
   b. ...and one-to-one (probably just a convenience function to group correspondences by from and select max to)
2. Geographical centroid (one-to-one) correspondences (I think there was discussion with ellisp / daviddiviny on this but not sure it got implemented?)
3. Geographical area overlap (both one-to-one and pro rata / one-to-many)

Definition of done:

• Lists of correspondences from various (all?) regions (and year-versions thereof) to various (all?) other regions (and year-versions)

Potential approaches:

• Just provide functions to do this, rather than pre-compute all combinations
• Pre-compute all combinations

Pros:

• I know I for one do at least one of the above over and over again for different projects, so having this in one place once would be super convenient
• Providing these pre-calculated options should, I think, cover the majority of use-cases (notably, supporting both one-to-one and one-to-many correspondences)

Cons:

• These aren't highly rigorous approaches and come with a bunch of caveats, obviously. Would need to include those in the documentation
• Pre-computing these for all (?) combinations of regions would take a long time

Happy to do the PR on this one, but interested in others' philosophical takes. @wfmackey ?

[mjkerrison]

A motivating factor I forgot to mention explicitly - the main reason for including this in {absmapsdata} rather than a separate package is just that 95% of the time I'm doing this with things from {absmapsdata}.

[mjkerrison]

The ABS have also published their 2021 correspondences so maybe worth updating those at the same time.

[wfmackey]

hi Michael - it would be great to have the self-calculated population-weighted correspondences.

I am wary of adding in self-calculated objects into this package, mainly because it muddies what is straight from the ABS and what has some non-ABS decision rules applied to it. But I think they would be great in a separate package that can be then called from strayr (as absmapsdata is now).

Keen to hear your thoughts