Implement a proposed modular API to specify a data (i.e. case) generating process

[SamuelBrand1]

This PR introduces the modular API to the data generating process discussed in #38 .

Major additions

Diagrammatically, the "modules" (in workflow rather than Julia sense) that are implemented are below. Dotted lines fill in the currently unimplemented steps.

Unit tests are in test scripts here.

A script to run in TestDev mode which demonstrates the currently implemented steps is here.

Minor additions

• Made the RW distributional test slightly more precise
• Moved the random walk default priors from a const object to a simple construction function

Missing feature

I have not implemented docstrings in some functions here ahead of addressing #35

flowchart TD

    A["Underlying dists.
---------------------
EpiData"]

    B["Choice of target 
for latent process
---------------------
DirectInfections
    ExpGrowthRate
    Renewal"]

C["Observational Data
---------------------
Obs. cases y_t"]
D["Latent processes
---------------------
Random Walk"] 
E[Turing model constructor]
F["Latent Process priors"]
G[Posterior draws]
H[Posterior checking]
I[Post-processing]
DataW[Data wrangling and QC]

    A --> EpiModel
    B --> EpiModel
    EpiModel -->E
    C-->E
    D-->|random_walk| E
F-->|default_rw_priors|E
E-.->|sample...NUTS...| G
G-.->H
H-.->I
DataW-.->C

Loading

Closes #38
Closes #15

[seabbs]

This looks really nice. It feels a little still like there is a mixture of styles going on here for the different parts of the model (as in the issue) it would be great to have a little more clarity on that.

I think we might be best places though to review this design as it stands and then do another round if issue/PR iteration?

I've only sight read for now so will circle back once I have looked over the code locally.

[seabbs]

I think this is good to go once issues have been resolved (i.e. moved to issues) as is. I do think there is a bit more conversation to be had about the precise modular construction of the model but I think we can iterate on this over time and should move on to postprocessing and formulating the different models likely first.