Changelog

Backend switched from rstan to cmdstanr. Models are now fit with CmdStanR and the latest CmdStan. This is faster, tracks upstream Stan, and removes the fragile install-time C++ compilation. The Stan programs are compiled on first use and cached (see compile_epidemia()).
Stan programs updated to modern Stan syntax (Stan 2.33+/CmdStan 2.36), including the new array declaration syntax (array[n] int x) and the vertical-bar CDF calls (normal_cdf(x | mu, sigma)). The external C++ csr_matrix_times_vector2 helper is now implemented in pure Stan.
Posterior draws are represented with the posterior package rather than an rstan stanfit object. as.matrix(), as.array(), as.data.frame(), summary(), posterior_rt(), posterior_predict() etc. behave as before.
rstanarm is no longer a dependency. The prior constructors (normal(), student_t(), cauchy(), exponential(), laplace(), lasso(), hs(), hs_plus(), product_normal(), lkj(), decov()) are now provided directly by epidemia, with identical behaviour. Existing code using these functions continues to work unchanged.
Speed: chains run in parallel by default (parallel_chains), and models are compiled once and cached.
No install-time compilation: the src/, configure and LinkingTo machinery has been removed.

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

Random walk terms parsed separately as input to stan files. Variance parameter sampled in stan, and so can make predictions.
pseudo-log scales for plot_infections and plot_observations
control over date range for all plots
option to plot smoothed Rt in plot_rt

epidemia 1.1.0