Likelihood for multi-type clusters

[adamkucharski]

It's also possible to define a likelihood for multi-type branching process for age-stratified data (Kucharski et al, PLOS Comp Biol, 2015, which is particularly relevant if the population introducing the infection is very different to that driving human-to-human transmission (e.g. older individuals more likely to be index cases for MERS-CoV).

There's a (not particularly tidy) implementation here: https://github.com/adamkucharski/subcritical_chains, which could also be combined with socialmixr.

If multi-type inference not a priority, it may be worth considering a multi-type simulation process for age-stratified social mixing data in meantime (currently implemented as simulate.data() in above repo.

[sbfnk]

That is a nice idea if a fairly substantial extension - would require some thinking about the best user interface.