Do you need to set a starting seed for repeatability?

[akeyel]

Describe the improvement that needs to be made

How repeatable are the MCMC model runs? I.e. would you get different results if you ran the model multiple times?
Would using set.seed() be honored by the underlying STAN code, or would it just get ignored?

[dylanhmorris]

The Stan PRNG seed is set separately from the R PRNG seed. Within the package, it is configured via get_mcmc_options
https://cdcgov.github.io/ww-inference-model/reference/get_mcmc_options.html

It strikes me that:

1. This should be better documented.
2. Currently, there is a fixed default (seed = 123). That should not be the case. I think I would advocate for default seed = NULL.
3. There's also the matter of seeding the random initial values. See R PRNG seeding when generate_random_inits = TRUE? #107.

[dylanhmorris]

This is addressed as of #175