A Basic Example
A Multilevel Model
First version submitted to CRAN
Bug fixed for latent infections and first Rt with pop_adjust
reorganized files to correctly attribute copyright.
New model for adding vaccination adjustments
latent infections switched to normal, from log-normal
Additional vignettes and model description
Additional noise options for infection process
Significantly more flexible modeling of seeding process
Many small bug fixes
Changes to general interface - new epiinf() function for representing infection model
Improved package website, with better description of the model and examples in the vignettes.
Ability to model latent infections explicitly - replacing renewal equation
Removed ‘pop’ and replaced with column of susceptibles in dataframe. This allows susceptible population to reduce over time due to vaccinations.
Improved error checking in epim(), with more informative messages
scaled_logit for epiobs
Full integration with Bayesplot package, and a plot.epimodel method which easily allows the user to choose different components of the model.
Fixed bug which meant “fullrank” was actually using “meanfield”
Ability to use random walks in the epiobs models
Choose between identity, scaled logit, and log link for epirt
Plot the (potentially transformed) linear predictors for both epiobs and epirt
Additional families for epiobs - normal and lognormal
Add summary method and printing for epimodel objects
Plots improved, allowing step for plot_rt, and improved formatting
Substantial changes to interface: Added epirt and epiobs objects
Different and more flexible observation models
Improved structure to epimodel objects
Refactoring of main epim function
Improved plots, including interactive plots using plotly
Forecast evaluation using coverage and different metrics
Ability to do an initial run fit to cumulatives within epim
Updated tests, documentation and vignettes
Improved model description in introduction vignette
Passes R CMD Check with no warnings
Updated installation instructions
Renamed stan files to avoid errors with Rstan 2.12.
Random walk terms parsed separately as input to stan files. Variance parameter sampled in stan, and so can make predictions.
pseudo-log scales for
control over date range for all plots
option to plot smoothed Rt in
Features for counterfactual analysis and predictions
Added vignette describing priors
Description of collinearity issues in ‘resolving problems’ vignette
Form for potential beta testers
Fixes to documentation in website
Separate index for website and github
Website and more extensive vignettes