Try this Pull Request!

Open Julia and type:

import Pkg
Pkg.activate(temp=true)
Pkg.add(url="https://github.com/CDCgov/Rt-without-renewal", rev="specific-priors", subdir="EpiAware")
using EpiAware

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.96%. Comparing base (0d4095e) to head (af6dfb6).

@@           Coverage Diff           @@
##             main     #565   +/-   ##
=======================================
  Coverage   90.96%   90.96%           
=======================================
  Files          60       60           
  Lines         863      863           
=======================================
  Hits          785      785           
  Misses         78       78           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Sorry I haven't reviewed the priors in detail. Will do so in the morning. The implantation seems correct but um yeah a bit horrifying.

I agree, the basic problem was that we identified:

• Diff AR is quite a different process to AR (obviously) because, approximately, its variance growth like $t$.
• ExpGrowthRate is quite different to other IGPs because its cumsum on the log-scale.
• DirectInfections on log-scale is therefore one less diff version of ExpGrowthRate.

Once I started reasoning on that you quite quickly end up with this horrible special case prior implementation. I couldn't think of a smooth dispatch, although maybe I just need to think a bit more.

Tidying this:

• Treating the default as one good choice at least reduces the bloat.
• Reasoning a bit more on how we expect the variance to be a priori and then reversing back towards a good choice of prior dists on the parameters would be much smoother.

Really good feedback.

Ongoing, I'm writing more logic into the doc strings :

This function sets the target standard deviation for an infection generating process (igp)
based on whether the latent process representation of its dynamics are stationary or non-stationary.

Stationary Processes

For Renewal process log(R_t) in the long run a fluctuation of 0.75 (e.g. ~ 75% of the mean) is not unexpected.
For Exponential Growth Rate process r_t in the long run a fluctuation of 0.1 is not unexpected e.g. going from
rt = 0.1 (7 day doubling time) to rt = -0.1 (7 day halving time) is a 2 x 0.1 fluctuation.
For Direct Infections process log(I_t) in the long run a fluctuation of 1.0 (i.e an order of magnitude) is not unexpected.

Non-Stationary Processes

For Renewal process log(R_t) in a single time srep run a fluctuation of 0.025 (e.g. ~ 2.5% of the mean) is not unexpected.
For Exponential Growth Rate process r_t in a single time step run a fluctuation of 0.005 is not unexpected.
For Direct Infections process log(I_t) in a single time step run a fluctuation of 0.025 is not unexpected.

And I'm not convinced the hard code version matches the scientific reasoning. Just doing something to check that before any commits

And I'm not convinced the hard code version matches the scientific reasoning. Just doing something to check that before any commits

To address this I've got #566 to merge into this branch. I think that this makes the logic much clearer, and therefore implements what we want to implement. Albeit it needs scientific review.