search

phac-nml-phrsd / ern

R library to estimate the Effective Reproduction Number (Rt)
Other
6 stars 1 forks source link

EpiNow2 paper follow up #214

Open seabbs opened 2 months ago

seabbs commented 2 months ago

Hi! Really looking forward to checking this out in detail!

I was just skimming the associated paper (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0305550) and saw this:

The package also provides tools for short-term forecasting of case reports but cannot handle explicitly non-daily (e.g., weekly) report

About EpiNow2. Unfortunately, it's not very well documented it can handle non-daily reporting processes. I think we spoke a little about this before when it was possible but required just throwing it at the model and fitting to it which we did a lot of during COVID-19 (not sure I would characterise that as requiring expert knowledge as per the table but hard to say!). Now it's a fully supported feature (https://epiforecasts.io/EpiNow2/reference/obs_opts.html - see the docs for the na arg). I'd be keen to see how this compares to what you are doing.

Sam

P.S I think this probably links to it being tricky to define what a software package can "do" when they all keep changing. I don't know what a good solution to this is aside from making sure to reference the explicit version being talked about but even then really hard.

davidchampredon commented 2 months ago

Thanks for the update @seabbs ! Apologies if we misinterpreted EpiNow2, we did this assessment based on the available documentation at the time. Fully agree with your comments about the trickiness of stating what packages can "do" at a given time...

seabbs commented 2 months ago

No worries at all. I think probably there needs to be some kind of living review/benchmark etc that lives somewhere independent and that people submit to. Would be a pretty cool collaboration.