epidemics 0.5.0

Breaking changes

1. Social contact matrices from socialmixr must no longer be transposed before use. model_default() and model_vacamole() now transpose the contact matrix internally, so matrices should be passed to population() exactly as socialmixr::contact_matrix() returns them (#259, @bahadzie). Existing code that calls t() on the contact matrix will continue to run without error, but will silently give incorrect results because the matrix is then transposed twice; remove the t() call when upgrading. Vignettes, examples, tests and the README have been updated accordingly.

2. The new_infections() argument compartments_from_susceptible has been renamed to exclude_compartments (#272, @bahadzie). Code that passes this argument by name must be updated.

Model functions

1. The compiled code underlying the ODE models has been migrated from Rcpp/C++ to odin (#263, @bahadzie, @alxsrobert). The 'default', 'Vacamole' and 'diphtheria' ODE systems are now declared as odin models under inst/odin/, odin has been added to Imports, and the previous C++ sources, their Rcpp bindings and the C++ linting workflow have been removed.

Classes

1. print() on a <population> object now also displays the initial conditions matrix (#269, @bahadzie).

Helper functions

1. Added the combine_populations() function to combine age-structured populations into a new population object (#262, @alxsrobert).

Bug fixes

1. Fixed model_default(), model_vacamole(), and model_diphtheria() silently dropping or merging demographic groups in their output when a model had 10 or more demographic groups (e.g. when combining four or more populations), caused by the demography group index being truncated to a single digit while reshaping model output (#278, @avallecam).

Documentation

1. Added a modelling-populations vignette on how to combine age-structured populations (#262, @alxsrobert).

2. Corrected the equations documented for model_vacamole() in the function documentation and the 'Modelling a two-dose vaccination campaign' vignette (#273, @bahadzie).

3. Corrected the vaccination rate used in the 'Modelling vaccination' vignette (#268, @bahadzie).

4. Corrected the Figure 1 caption in the 'Modelling multiple interventions' vignette (#270, @bahadzie).

5. Fixed rendering of equations on the pkgdown website for pkgdown >= 2.1.0 (#249, @Bisaloo).

6. Fixed broken links in the README (#293, @joshwlambert).

7. Updated the documentation of the internal argument-preparation helpers for the 'Vacamole' and 'diphtheria' models (#274, @bahadzie).

8. Set the pkgdown website development: mode to unreleased so that the single website matches the development version of the package installed by most users, and added website favicons (#275, @joshwlambert).

Package

1. Added dependabot.yml to .github/ to automate updating GitHub actions workflow versions (#283, @joshwlambert).

2. The minimum required R version is now 4.1.0, as the package's examples, vignettes and tests use the base R pipe (|>).

3. Updated calls to socialmixr::contact_matrix() to use the renamed age_limits argument, replacing the defunct age.limits, and accepted the resulting snapshot changes (#280, @avallecam).

4. Fixed the codecov/codecov-action variables in the test coverage workflow (#289, @joshwlambert), added persist-credentials: false to fix the update-citation-cff workflow (#290, @joshwlambert), and removed an unused workflow inherited from the package template (#244, @pratikunterwegs).

5. Updated all uses of socialmixr across tests, vignettes, examples and the README to pass survey_pop explicitly to socialmixr::contact_matrix(), removing deprecation warnings introduced in socialmixr 0.6.0; the minimum socialmixr version is now 0.6.0 (#281, @joshwlambert). Demography data is now taken from the wpp2024 package (added to Suggests and Remotes, installed from GitHub) rather than from the deprecated socialmixr::survey_country_population(), which is backed by the outdated wpp2017 data. The population year is 2006, matching the year in which the socialmixr POLYMOD participants were surveyed; previously socialmixr used 2005, as wpp2017 only provides population estimates at five-year intervals. Demography vectors and symmetric contact matrices in examples, vignettes and tests therefore change slightly (by up to ~1%) relative to previous releases, reflecting the revised UN estimates for 2006.