search

nfidd / nfidd

Course on nowcasting and forecasting of infectious disease dynamcis
https://nfidd.github.io/nfidd/
MIT License
2 stars 0 forks source link

split learning objectives into sessions #15

Closed sbfnk closed 3 weeks ago

sbfnk commented 1 month ago

~This is already superseded by #14 but were my initial thoughts for splitting it into 5 session of 3 hours. To be closed and replaced by a PR that introduces a schdule.~ superseded below

sbfnk commented 4 weeks ago

Now replaced by a tentative (tight!) schedule. We could and perhaps should still consider cutting some material.

sbfnk commented 4 weeks ago

Will need to update to https://github.com/nfidd/nfidd/pull/14#issuecomment-2092784489 - will do following review (which I suspect may end up in restructuring anyway).

seabbs commented 4 weeks ago

(tight!) schedule.

Currently feels a bit like Hermoine's class schedule from book 3 (due to the time travel)

kathsherratt commented 4 weeks ago

I'd be a bit tempted to start straight off the bat on Monday with the short intro to Rt estimation. Then widen out into delays in general and nowcasting. People will be familiar with Rt, and I think it can be good to start with that in order to introduce the general idea of more carefully thinking through and estimating latent/unobserved processes. I'm not very attached to this suggestion and can see the logic of how it is now, so feel free to ignore.

seabbs commented 4 weeks ago

So my main push back would be that the models for renewal equations are much more complicated to write down/read vs a simple distribution and so it wold be quite a hot start to the process. I would also push back a bit that people won't be familiar with say an incubation period.

I also like the idea of people know about delays ahead of Rt estimation because then right when you start talking about Rt you can make it clear its a latent quantity and its latent because of these delays you have just introduced.

On the other hand if splitting Rt into an underlying latent process lecture (i.e just the renewal) and then a "real-world" rt (w/ delays) after the delay bit I can see it working out.

kathsherratt commented 4 weeks ago

Yeah that's all fair! Keep as is then :)