Full package review for {bpmodels} CRAN release

[jamesmbaazam]

This is a full package review in preparation for submitting {bpmodels} to CRAN. It would be helpful to provide comments that help resolve the checklist provided here #80.

Modalities

How to become a full reviewer

Self-assignment under the "Reviewers" tab on the right side of this window.

Required number of reviewers

Maximum of 2. PS: Don't self-assign when this is reached.

Full reviewers

1. @joshwlambert
2. @pratikunterwegs

Deadline

Not currently set.

What next?

• This PR will be closed when all relevant issues have been resolved.
• Closing this PR will not close any issues raised but will serve as a reference for raising issues to be resolved in subsequent pull requests.

[jamesmbaazam]

• I would reword "The negative binomial distribution is commonly used in epidemiology to account for individual variation in transmissibility, also known as superspreading". As small amounts of variation (e.g. poisson or geometric) are not considered superspreading.

Not sure I understand how your second statement relates to the first.

[jamesmbaazam]

• It seems that in figure 1 of the COVID vignette many of the grey lines are above the red median line. I would have expected the red line to sit in the middle of each stochastic simulation?

That's a good point. I'll crosscheck that.

[jamesmbaazam]

• It might be good to have a control on simulating negative serial intervals. For now you could limit the functionality to work with strictly positive serial interval.

I am not sure that this is a necessary feature as we do not want to restrict the user on what types of serial intervals to simulate. There's a case for users wanting to simulate negative serial intervals to pursue questions about the infector being infectious before developing symptoms (See the "serial interval" section of this paper that for example mentions that observed serial intervals can be negative, and [this one](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7239082/#:~:text=Negative%20serial%20intervals%20(left%20of,asymptomatic%20or%20mildly%20symptomatic%20cases) that observed negative serial intervals, and other examples in the literature.

@sbfnk Thoughts on this requested feature?

[jamesmbaazam]

• The code chunk

chain_sizes <- c(1, 1, 4, 7) # example of observed chain sizes
chain_ll(chain_sizes, "binom", "size", size = 1, prob = 0.5,
         nsim_offspring = 100)

returns -Inf is this correct?
Could it be because you're using the binomial instead of the negative binomial?

There is an analytical solution for the negative binomial size distribution. This code chunk demos the simulated binomial size result.

[sbfnk]

• In the get started vignette, is a vector of four chains sizes a good example to use? I don't know, but I would imagine the power to differentiate between generating models is low when the sample size is four (assuming that the sample size is the number of transmission chains and not the total size of all chains)

It's a good point -

• It might be good to have a control on simulating negative serial intervals. For now you could limit the functionality to work with strictly positive serial interval.

I am not sure that this is a necessary feature as we do not want to restrict the user on what types of serial intervals to simulate. There's a case for users wanting to simulate negative serial intervals to pursue questions about the infector being infectious before developing symptoms (See the "serial interval" section of this paper that for example mentions that observed serial intervals can be negative, and [this one](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7239082/#:~:text=Negative%20serial%20intervals%20(left%20of,asymptomatic%20or%20mildly%20symptomatic%20cases) that observed negative serial intervals, and other examples in the literature.

@sbfnk Thoughts on this requested feature?

I agree - supporting negative SIs is a feature not a bug.

[jamesmbaazam]

All issues raised through this full package review have been addressed and logged as GitHub issues, so I am closing this PR now.