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
Want to allow users to still fit the model without a distance matrix/an assumed spatial correlation function (defaults to independence). For now, I think we can treat them as fully iid variables, but eventually we will want to explore fitting a correlation matrix without distance data using the MVN structure (and then we can compare all 3 approaches + additional spatial functions)
Requirements
[ ] be able to fit a model without distance data, to replicate the main wwinference model
[ ] user friendly interface for toggling on and off the spatial correlation function component of the model
Out of scope
[ ] fit correlation matrix to MVN without spatial data (think this should be a separate PR, but we want to keep this functionality in mind when building out this feature)
Goal
Want to allow users to still fit the model without a distance matrix/an assumed spatial correlation function (defaults to independence). For now, I think we can treat them as fully iid variables, but eventually we will want to explore fitting a correlation matrix without distance data using the MVN structure (and then we can compare all 3 approaches + additional spatial functions)
Requirements
wwinferencemodelOut of scope