Comment does not match code in `outbreak_setup()`

[joshwlambert]

The comment on L47 of outbreak_setup() states:

# set isolation time for cluster to minimum time of onset of symptoms + draw from delay distribution

however, the code on L48 does not use the minimum time, instead it uses the vector of onset times.

[pearsonca]

Revisiting the methods in paper:

We initialised the branching process with five, 20, or 40 cases to represent a newly detected outbreak of varying size. Initial symptomatic cases were then isolated after symptom onset with a delay drawn from the onset-to-isolation distribution ...

Rewinding the repo to look at the (second) submission state, no changes to this element.

Conceptually, the pmin version here would correspond to:

• the new case cluster all derives from a single exposure event (compatible with giving them all the same exposure time)
• they are all known as exposed somehow (inconsistent with subsequent tracing etc governed by input parameters)
• but there's no acting on that exposure until the first of them shows symptoms

That ...doesn't seem like a reasonable initial condition, especially when the analysis explores varying initial infections independently of natural history parameters. The transmission chains are also non-interacting, so to me it seems like the initial size is more about ensuring that one doesn't observe stochastic burnout, not some notion of the initial cases actually representing a cluster.

I recommend replace the comment, instead indicating that the intent is simply to have an initial set of independent cases.

My guess is that having a distribution in isolation times should make increasing the number of initial cases more reliably preclude early burnout (when statistically unlikely)

[joshwlambert]

Thanks for looking into this and for the detailed description @pearsonca.

Would you mind explaining what you mean by:

they are all known as exposed somehow

I'm not sure I understand as, all initial cases are missed and a proportion are asymptomatic, thus unknown to the surveillance system until symptom onset.

[pearsonca]

Sure; i mean that giving them all the same isolation time could represent that they had some known exposing incident (which definitionally as you point out, they do not). As in, I know a bunch of people all got exposed at some event - i find that out at some time (via passive detection of one of those people, which is their onset + isolation delay) and then (instaneously) i apply isolation to all the other people in that group.

Seems like a weird assumption, so I'm guessing the comment is wrong and we should just delete it and continue independently sampling them.

[joshwlambert]

This issue was resolved when PR #173 was merged.

Thanks @pearsonca for spotting the original mistake, the detailed description, and for looking into the version history of the code and corresponding information in the paper.