search

CDCgov / ww-inference-model

An in-development R package and a Bayesian hierarchical model jointly fitting multiple "local" wastewater data streams and "global" case count data to produce nowcasts and forecasts of both observations
https://cdcgov.github.io/ww-inference-model/
Apache License 2.0
16 stars 2 forks source link

Review initialization approach, document it more, and possibly make it more generic #66

Open dylanhmorris opened 2 months ago

dylanhmorris commented 2 months ago

I think this documentation needs rewording for clarity. If the goal of this function is to initialize each parameter near its prior mode, should say that explicitly, and also clarify that the distributional shapes specified below are the prior distribution shapes from the actual model. But it's not clear to me that that's exactly true in all cases.

Also, if the goal is to be less diffuse than the priors, why not provide a multiplicative factor for transforming the prior sds, rather than hard coding, i.e. something like

my_scalar_param <- rnorm(1, params$my_param_prior_mean, scaling_factor * params$my_param_prior_sd)

_Originally posted by @dylanhmorris in https://github.com/CDCgov/ww-inference-model/pull/54#discussion_r1714363453_