Fit negative binomial offspring distribution

[adamkucharski]

Add functionality to estimate R and k from individual-level transmission chain data (e.g. Adam et al, 2020. This functionality could also be integrated with readepi and godataR to allow direct estimation from imported timeseries. The values could also be passed to bpmodels to allow for simulation of branching process outbreaks from the estimated values.

[adamkucharski]

The fitting can be done with fitdistrplus::fitdist(), so main task will be data formatting for input/output (e.g. from contact datasets) to reduce need for manual standardisation

[joshwlambert]

@adamkucharski given, as you say above, that this functionality already exists in packages like {fitdistrplus} I wonder whether the best way to tackle this issue is with a vignette. It could be an elaboration on the example given in the readme?

What do you think?

Thoughts on vignette structure could be:

1. Read in and tidy data
2. Fit negative binomial model with {fitdistrplus}
3. Apply summary metrics from {superspreading}
4. Simulate with {bpmodels} with negative binomial parameters

[adamkucharski]

I agree, that sounds like a good structure for a vignette.

[joshwlambert]

PR #15 tackles this issue, but not with the same structure outlined above. Instead it covers fitting models with {fitdistrplus}.

Instead of extending this vignette (which also tackles #4), I propose I open a new issue which requests an additional vignette which will cover: reading in and tidy data, applying summary metrics from {superspreading}, and simulating with {bpmodels} with negative binomial parameters.