Adding a new package to epinowcast github

Hi everyone,

My name is Daniel (or Sang Woo), and I’m a 4th year PhD student at Princeton University (see intro here: Sang Woo Park - intro).

@samabbott and I have been working on a paper investigating different methods for estimating delay distributions while accounting for truncation and censoring biases. We also present a new method of fitting to backward delay distributions, which is closely linked to the nowcasting problem. In doing so, we (mostly Sam) have been developing it into an R package (GitHub - parksw3/epidist-paper).

We thought adding the package to the epinowcast github repository would be helpful for its future development and increasing the overall community engagement, given that fitting delay distribution is an extremely common problem in analyzing outbreak data. So we wanted to ask what other people thought of it.

Can we add epidist to the epinowcast github repo?

  • Add
  • Don’t add

0 voters

We welcome any other feedback.

Thanks!

1 Like

We haven’t discussed our onboarding for new packages yet and its definitely something we need to address and formalise going forward. Perhaps we can cover ways to do this at the next community meeting.

Personally, I really like the Ropensci model which involves software review, a clear need, and then ongoing support such as finding maintainers etc. Given we are a bit more integrated we might need to slightly tweak this approach but it might be a good starting point nonetheless.

I think there is a strong case for this package as, as @sangwoopark says, fitting distributions is very common during outbreaks. I also think our implementation, which extends the brms R package, is a really clean one that promises a lot of additional functionality in the future for very little cost. Whilst it has some overlap with the core epinowcast model there is a fairly clear use case and I can imagine quite a few scenarios where they might interact as well (for example estimating delays as inputs for latent parts of the epinowcast model).

As part of this work we have looked across the R ecosystem and there isn’t really much we have found that covers the core issues of estimating delays during outbreaks so I think there is a strong case for a new package rather than trying to add features to any of the current options.

So far this is looking like a strong yes so assuming nothing change will go ahead and port in and then post an update.

1 Like