0.2.0 package pre-release

We have nearly implemented the planned functionality for the 0.2.0 release (see here for the plan: Changelog • epinowcast and here for currently added features: Changelog • epinowcast). There is some really exciting stuff in here and a lot of work from many people so really keen to get this out in the world.

I’ve made a PR (see 0.2.0 Release by seabbs · Pull Request #180 · epinowcast/epinowcast · GitHub) with the list of issues I think we need to deal with before we can make a release. It would be great if people can take a look/comment/add/remove etc. This also means that we should start planning which issues to address in 0.3.0. For me, the key things to do are stabilise the interface (i.e Use of enw_ prefix and Add data format converters · Issue #144 · epinowcast/epinowcast · GitHub), decide on and implement a definition for the max delay and get forecasting functionality implemented (Add future forecasting ability · Issue #3 · epinowcast/epinowcast · GitHub). If possible I would also love to add non-parametric delays (Add non-parametric reference date model · Issue #4 · epinowcast/epinowcast · GitHub) and decoupling of the expectation/reporting delay model (i.e so you can fit conditional distributions + use the expectation model on non-right truncated data - no issue for this one but we need it).

There are lots of other features we likely want. I’d really like to see people suggest some in the issues etc :slight_smile:

In terms of where we should put the planning for 0.3.0 I am not that sure? We could make a discuss here for now and then move it to GitHub once things are more sorted out or open a meta-issue on GitHub.

Thanks for compiling the pre-release @samabbott ! I have commented on 0.2.0 on github.

Regarding 0.3.0 I think it would be nice to have the discussion here first, making it more accessible and visible hopefully.

For me, some valuable points in addition to the ones mentioned by @samabbott would be

  • investing in documentation and more/better vignettes
  • adding basic simulation capabilities (potentially as separate package)
  • plotting estimated reporting delays and share of missing reference dates
  • extending ability to specify priors
1 Like

I agree with all of those suggestions. I also think we need to think more about the user journey and make sure the package is relatively friendly to the various kinds of data that people usually have.

We now have a candidate release on main which really needs its tires kicking (see draft release notes here: Changelog • epinowcast).

In particular, could people check the news item and make sure I haven’t missed anything or anyone? We are also really really crying out for a model schematic as currently we are just trying to :exploding_head: peoples brains I think.

I’ve also set up a meta-tracking issue for this release with some final things to do. I am going to propose we release it by writing a blog post on the epinowcast.org site which I will draft up and then post here (perhaps over the weekend or next week).

Something to note is that in fixing our R universe I have updated it so that it builds our latest release vs the latest code on main. This gives us a three-level release process with manual production releases via the GitHub release framework, a stable development version (i.e main), and an unstable development version (develop). This should let us have a bit more confidence about pushing out small releases rapidly.

Release and news look good from my side!

Fully agree that we need a model figure, maybe even two versions with different complexity. Maybe something to draft conceptually during the next meeting and then someone can produce a decent figure based on it.

Nice one with the R universe, so the idea is we can be more ruthless on develop? :smiley:

Good idea to have a brainstorm at the next meeting. Also worth discussing the kind of style we want to go for (i.e how much math).

Nice one with the R universe, so the idea is we can be more ruthless on develop ? :smiley:

Yes exactly and have a working development version on main.

Made the release offical as we seem happy with it and pushed out a twoot: Sam Abbott: "Epinowcast #rstats 📦 release 0.2.0 is now out (as…" - Fosstodon

Still drafting a release post so will share that once I have something!