Add simulated data from Gostic, 2020 for benchmarking

[zsusswein]

This commit re-uses the data processing and documentation from CDCgov/cfa-epinow2-pipeline#17. That repo is open source and public domain by nature of being USG property. I think it's convenient to re-use, but @athowes and @kgostic there's a little text in there that you both suggested, so I'd appreciate if you could give your permission for the re-use here! I've made sure to credit you both in the commit as it's partially your writing.

It's probably best practice to make the data prep and processing here fully reproducible in data-raw/ but it's quite a pain to do, with a mixture of shells cripting and R scripting needed. I skipped it out of convenience, but if you have a really strong feeling that it's required @seabbs, let me know and we can revisit. I did it for the generation interval PMF but not the dataset recreation.

Closes #9

Co-authored-by: Adam Howes [email protected]
Co-authored-by: Katie Gostic [email protected]

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Additional details and impacted files

📢 Thoughts on this report? Let us know!

[seabbs]

I think its reasonable to skip the toy rt data generating scripts but ideally I would like to the exact code that produced it if its available (i.e the CFA code?). I also think that the code for the pmf should be very minimal so maybe include.

[zsusswein]

ideally I would like to the exact code that produced it

I'll do it with primarycensoreddist. How stable is that API?

[athowes]

I agree with Sam about might as well document going from the data to the PMF in data-raw if that isn't already something you've got. Also perhaps move to primarycensordist

[seabbs]

I'll do it with primarycensoreddist. How stable is that API?

For the R code its stable unless I receive any feedback with strong views

[seabbs]

Looks good to me. Remaining conversation points are really questions as to whether features here should actually be part of primarycensorreddist (i.e do they have a wider application).

[kgostic]

I really am squirming seeing my name all over this and would like it if you could call the dataset something like "sudden-Rt-change" but this isn't the hill I'm going to die on if it makes a bunch of work for you.

Agree too that it would be ideal to be able to re-create the dataset and think that's a lift we could tackle in the future, but think it's ok to remain an issue in the backlog for now!