Merge branch 'main' into vectorise-spectral-density

NEWS.md

@@ -26,6 +26,7 @@
 - a bug was fixed where `forecast_secondary()` did not work with fixed delays. By @sbfnk in #717 and reviewed by @seabbs.
 - a bug was fixed that caused delay option functions to report an error if only the tolerance was specified. By @sbfnk.
 - a bug was fixed that led to the truncation PMF being shortened from the wrong side when the truncation PMF was longer than the supplied data. By @seabbs in #736 and reviewed by @sbfnk and @jamesmbaazam.
+- a bug was fixed that caused internal validation checks on delay distributions to fail if they contained non-parametric distributions. By @jamesmbaazam in #750 and reviewed by @seabbs.
 
 ## Documentation
 

R/checks.R

@@ -81,12 +81,14 @@ check_stan_delay <- function(dist) {
   # Check that `dist` has parameters that are either numeric or normal
   # distributions with numeric parameters and infinite maximum
   numeric_or_normal <- unlist(lapply(seq_len(ndist(dist)), function(id) {
-    params <- get_parameters(dist, id)
-    vapply(params, function(x) {
-      is.numeric(x) ||
-        (is(x, "dist_spec") && get_distribution(x) == "normal" &&
-           is.infinite(max(x)))
-    }, logical(1))
+    if (get_distribution(dist, id) != "nonparametric") {
+      params <- get_parameters(dist, id)
+      vapply(params, function(x) {
+        is.numeric(x) ||
+          (is(x, "dist_spec") && get_distribution(x) == "normal" &&
+             is.infinite(max(x)))
+      }, logical(1))
+    }
   }))
   if (!all(numeric_or_normal)) {
     stop(

README.md

@@ -90,6 +90,20 @@ cost of the granularity of estimates or real-time performance, include:
 - Adjustment for the remaining susceptible population beyond the
   forecast horizon.
 
+By default, all these models are fit with [MCMC
+sampling](https://mc-stan.org/docs/reference-manual/mcmc.html) using the
+[`rstan`](https://mc-stan.org/users/interfaces/rstan) R package as the
+backend. Users can, however, switch to use approximate algorithms like
+[variational
+inference](https://en.wikipedia.org/wiki/Variational_Bayesian_methods),
+the
+[pathfinder](https://mc-stan.org/docs/reference-manual/pathfinder.html)
+algorithm, or [Laplace
+approximation](https://mc-stan.org/docs/reference-manual/laplace.html)
+especially for quick prototyping. The latter two methods are provided
+through the [`cmdstanr`](https://mc-stan.org/cmdstanr/) R package, so
+users will have to install that separately.
+
 The documentation for `estimate_infections` provides examples of the
 implementation of the different options available.
 
@@ -175,6 +189,12 @@ the two main functions in the package and how to set up them up. It also
 discusses how to summarise and visualise the results after running the
 models.
 
+More broadly, users can also learn the details of estimating delay
+distributions, nowcasting, and forecasting in a structured way through
+the free and open short-course, [“Nowcasting and forecasting infectious
+disease dynamics”](https://nfidd.github.io/nfidd/), developed by some
+authors of this package.
+
 </details>
 <details>
 <summary>