Joint estimation of R and k

[adamkucharski]

It would be useful to have functionality to jointly estimate R and k using the functions in likelihoods.R. Perhaps as a starting point by using simple profile likelihood methods similar to the ones here (which also includes some chain data for MERS): https://github.com/calthaus/MERS

[sbfnk]

In principle this functionality already exists (though maybe needs to be surfaced a bit more) by wrapping chain_ll with a negative binomial and two parameters into an estimator / MCMC / etc.

[jamesmbaazam]

@adamkucharski I believe this issue will be solved by quickfit and should be closed??

[adamkucharski]

Reopening as based on current discussions, sounds like would be useful to include as a vignette in epichains. I've got loads of cluster size data kicking around from other work if useful (e.g. MERS, H5N1, H7N9).

More discussion of functionality for estimation (e.g. normal approx, marginal, profile likelihood) is in this issue on quickfit. Also tagging @joshwlambert

[jamesmbaazam]

Thanks @adamkucharski. Will transfer this issue to {epichains}.

[jamesmbaazam]

The suggestion to include estimation functionality in {epichains} has been discussed a few times and deemed out of the scope of this package as it might introduce heavy dependencies and potentially become a maintenance burden. It might also require being opinionated about which frameworks to use. There has been some discussion of the pros and cons of existing options in https://github.com/orgs/epiverse-trace/discussions/185 and will be useful to anyone interested in pursuing this further.

The vignette suggested here will be addressed when epiverse-trace/howto#57 is merged into the how-to guides. Users who are interested in this should explore the options provided there.

In light of all this, I am closing this issue.