sempwn / cr0eso Goto Github PK

Would be better to import ... as a function parameter that can be passed into rstan::sampling

For more flexible modeling, would be useful to include a covariate matrix for locations and predictors that can be used to explain variance of R0 for each location.

For example, we may have difference facility types we would want to use to adjust R0 estimates, or type of ventilation, age of building, density etc.

https://deanattali.com/blog/migrating-travis-to-github/

• Before seir_model_fit runs add in some basic data checks like whether there are missing values (there shouldn't be).
• If there are a large number of zero cases which should throw a warning that may lead to convergence issues.
• Add in functionality to plot data before model fitting.

Add ability to change priors within model

Stan code should get updated so that it generates posterior predictive distribution for cases, as well as for the counterfactual where there is no intervention.

With these generated posteriors, can then use the tidybayes package to extract and plot. This can also help to tidy up the hom_plot_incidence_by_location() function.

Create function to use rstan fit object and run a counterfactual scenario for each location where no intervention takes place.

Output should be a class with a print and plot function and provide outputs in tidy format.

Follow guidlines here: https://r-pkgs.org/data.html

See no visible global function definition warnings on R CMD check.

Also, I wouldn't import the tidyverse, just the specific packages you need.

Don't forget the stats imports: "quantile", "rpois"

Example currently referencing an uninstantiated object. Easy fix would be to save a model fit and use that to generate posterior plots

Create function that has a list of model outputs as a list and returns a table of marginal posterior medians and 95% CrI for R0, zeta, and critical time