Index cases in outbreak_setup() are now only isolated if they test positive (using the test_sensitivity parameter), as opposed to previously where all symptomatic index cases were isolated. The updated outbreak_setup() matches the testing pathway in outbreak_step(). outbreak_setup() gains an interventions argument. Addresses #231 by @joshwlambert in #232 and reviewed by @pearsonca.
Interventions can now activate after a delay rather than being active from the start of the outbreak. The symptomatic_traced argument of event_prob_opts() and the test_sensitivity argument of intervention_opts() can now accept a function of time returning probabilities as well as a numeric scalar (constant over the simulation, as before). The function is evaluated at the contact's exposure time (symptomatic_traced) or the case's symptom onset time (test_sensitivity) in outbreak_step(), so contact tracing and testing can vary over time. A new internal as_prob_function() coerces scalar inputs to constant functions. Addresses #229 by @joshwlambert in #230 and reviewed by @pearsonca.
The time-based stopping criterion in outbreak_continue() now uses the earliest exposure time of unprocessed cases rather than the latest symptom onset, so long incubation periods no longer end the simulation prematurely and changing cap_max_days no longer alters in-window case counts. Addresses #163 and #225 by @joshwlambert in #228 and reviewed by @pearsonca.
Symptomatic individuals can now self-isolate without requiring a positive test. event_prob_opts() gains a symptomatic_self_isolate argument (the proportion of symptomatic cases that self-isolate, default 0) and delay_opts() gains an onset_to_self_isolation argument, (by default a function generating Inf so that no individual self-isolates). A new internal cross_check_opts() checks that the delay_opts() and event_prob_opts() settings are mutually compatible.
In outbreak_step() a case's isolation time is now the earliest of three pathways: self-isolation, a positive test, and contact tracing. Tracing under quarantine is now exposure-based, so all traced contacts, including asymptomatic ones, are isolated when their infector is isolated regardless of symptom status (previously asymptomatic cases were never isolated); sample_offspring() reduces post-isolation transmission for asymptomatic cases accordingly. Addresses #212 and #227 by @joshwlambert in #213 and reviewed by @pearsonca and @sbfnk.
The package test suite has been improved, adding several new unit tests, and increasing the test coverage to 100%. The {testthat} edition is incremented to use the 3rd edition. Snapshot (regression) tests are added for outbreak_setup() and scenario_sim(). Addresses #100 by @joshwlambert in #160 and reviewed by @pearsonca.
Updated package maintainer to @joshwlambert. By @joshwlambert in #223 and reviewed by @pearsonca and @sbfnk.
Added vignettes on:
ringbp.Rmd), containing an overview of the {ringbp} model and how to parameterise it, with an explanation of the model parameters, as well as how to plot the outbreak, and summarise the results (e.g. extinct_prob()). There is also a simplified version of the Hellewell et al. analysis. Addresses #139 by @joshwlambert in #201 and reviewed by @sbfnk and @pearsonca.ringbp-model.Rmd), containing model schematic figures which describe and illustrate how disease transmission and interventions (isolation, contact tracing and quarantine) are structured in the {ringbp} model. The vignette also links the epidemiological parameters that control the interventions, disease transmissibility and delay distributions with the R code for parameterising a {ringbp} simulation. Addresses #140 by @joshwlambert in #188 and reviewed by @sbfnk and @pearsonca.{ringbp} across a parameter set (parameter-sweep.Rmd) replaces the parameter_sweep() function. Addresses #86, #125 by @joshwlambert in #127 and reviewed by @pearsonca and @sbfnk.Added CONTRIBUTING.md. Addresses #109 and #203 by @joshwlambert in #199 and reviewed by @pearsonca.
Improved weekly aggregation in outbreak_model(). Addresses #168 by @joshwlambert in #197 and reviewed by @sbfnk.
Added test sensitivity to outbreak model (#176) and renamed symptomatic_ascertained to symptomatic_traced (#208) by @joshwlambert in #196 and reviewed by @sbfnk and @pearsonca.
The package test suite has been improved, adding several new unit tests, and increasing the test coverage to 100%. The {testthat} edition is incremented to use the 3rd edition. Snapshot (regression) tests are added for outbreak_setup() and scenario_sim(). Addresses #100, #178, #181 by @joshwlambert in #160 and #182 and reviewed by @pearsonca (#160) and @sbfnk (#182).
Fixed per-generation offspring distribution sampling. Previously, offspring counts were redrawn for every infector at every call to outbreak_step(), so the same infector could generate offspring across multiple generations. Offspring are now drawn once per infector via a new internal sample_offspring() function, and the isolated column in case_data is renamed to sampled to reflect that the infector has had their chance to infect. new_cases is now an integer. Because of the bug fix, users should expect simulation results to change. Addresses #148 by @joshwlambert in #169 and reviewed by @sbfnk.
Enhanced interaction between simulation and extinction functionality:
outbreak_model() and scenario_sim() return the extinction status as an attribute (extinct) of the data.table. This removes the need to calculate extinction from weekly case data in extinct_prob() and detect_extinct(). Extinction calculation code still exists in extinct_prob() and detect_extinct() for backwards compatibility and if the attribute is dropped.scenario_sim() returns a cap_cases attribute with the data.table and the cap_cases argument is removed from extinct_prob() and detect_extinct().outbreak_df_week argument in extinct_prob() and detect_extinct() is renamed to scenario and the week_range argument is renamed to extinction_week. The extinction_week argument can now accept: 1) an integer scalar, 2) an integer vector of length 2 with the lower and upper bounds of the extinction window, and 3) an integer vector of length n. The integer scalar is used as a cutoff for determining if extinction has occurred by the cutoff. The integer vectors are used as extinction windows, if users want to examine if extinction has occurred before a time that is not until the end of the outbreak. The new default for extinction_week argument is max(scenario$week) - 1 which means the last 2 weeks of the outbreak are used as the extinction window.detect_extinct() to ensure the extinction_week is within the simulated outbreak.Addresses #132 by @joshwlambert in #161 and reviewed by @pearsonca and @sbfnk.
Upgraded the pkgdown website to Bootstrap v5. Addresses #131 by @joshwlambert in #143 and reviewed by @pearsonca.
The lower bound for the generation time returned by incubation_to_generation_time() is now greater or equal to the latent period (default 0) instead of 1. A latent_period argument has been added to delay_opts(). incubation_to_generation_time() gains two new arguments: exposure_time and latent_period to prevent a bug where an infectee's exposure precedes an infector's exposure time. When latent_period > 0, generation times are left‑truncated at latent_period; this can reduce the realised presymptomatic transmission proportion, for which a warning is issued and the realised value is reported after simulation. Addresses #124 by @joshwlambert in #142 and reviewed by @sbfnk and @pearsonca.
The outbreak_continue() function is added to house the simulation loop logic. Addresses #133 by @joshwlambert in #150 and reviewed by @pearsonca and @sbfnk.
The outbreak simulation functions (scenario_sim(), outbreak_model(), outbreak_setup() and outbreak_step()) have been refactored to provide a more modular and functional interface. New delay_opts(), event_prob_opts(), intervention_opts(), offspring_opts(), and sim_opts() helper functions are added. check_dist_func() is added. The parameter_sweep() function is removed and converted into a vignette ({purrr} is removed as a package dependency). prop_presymptomatic_to_alpha() is renamed to presymptomatic_transmission_to_alpha(). Addresses #65, #91 by @joshwlambert in #127 and reviewed by @pearsonca and @sbfnk.
The inf_fn() function has been renamed to incubation_to_generation_time() and the k function argument (scenario_sim(), outbreak_model(), outbreak_step()) has been renamed prop_presymptomatic (in scenario_sim() and outbreak_model()) or alpha (in outbreak_step()). The internal prop_presymptomatic_to_alpha() function has been added to do the conversion from prop_presymptomatic to alpha. Addresses #119, #120 by @joshwlambert in #123 and reviewed by @pearsonca and @sbfnk.
Added input checking, removed unused argument defaults, consistently ordered arguments in functions, and moved function argument documentation to functions that use the argument. Addresses #89, #91, #93, #116 by @joshwlambert in #117 and reviewed by @pearsonca and @sbfnk.
Function argument and variable names are consistently named and styled (snake case). Arguments that were abbreviated (e.g. iso) now use the full word, and arguments that had aliases (e.g. prop_ascertain and control_effectiveness) now use a single name. By @joshwlambert in #112 and reviewed by @pearsonca and @sbfnk.
Fixed a bug in the implementation of quarantine, where isolation could happen later than onset + delay, if this was earlier than the isolation time of the infector. By @sbfnk in #107 and reviewed by @pearsonca and @joshwlambert.
Changed internal sampling to vector draws in several locations; in some cases, this was an optimization, but in others it was a bug fix correcting the use of single draw where there should have been many. Also incorporates other vectorization changes to improve performance. Because of the bug fix, users should expect results to change. Addresses issues #90, #92, #94, in #98 by @pearsonca (now added as a contributor) with review by @sbfnk and @joshwlambert.
The minimum R version required by the {ringbp} package is now >= 4.4.0. This is due to the dependency on {Matrix}. A GitHub actions workflow for R CMD check has been added to check the package is valid on the minimum required R version. Addressed in #85 by @joshwlambert with review by @sbfnk.
Parameterisation of delay distributions accepted by the model have been generalised in scenario_sim() and outbreak_model(). delay_shape and delay_scale arguments have been replaced by the onset_to_isolation argument, and an incubation_period argument is added. Addresses issues #59 and #63 in #84 by @joshwlambert with review by @sbfnk and @pearsonca.
Function examples are added and updated, and \dontrun{} is removed. Addresses issue #27 in #83 by @joshwlambert with review by @pearsonca.
Use {roxyglobals} to manage global variables. Addresses issue #66 in #81 by @joshwlambert with review by @sbfnk.
Remove silent printing of returned data.tables from outbreak_model() and scenario_sim() by adding [] to return(). Addresses issue #64 in #80 by @joshwlambert with review from @sbfnk.
Function documentation is updated, including function arguments, function outputs, inheriting arguments across functions and fixing out-of-date references. Addressed issue #58 and #67 in #78 by @joshwlambert with review by @sbfnk and @pearsonca and issue #79 in #82 by @joshwlambert with review by @sbfnk.
Added unit tests. By @timcdlucas in #43 and reviewed by @sbfnk.
Added testing and checking infrastructure using github actions.