Add estimate_dist() with primarycensored support

[sbfnk]

Description

This PR closes #350.

Adds estimate_dist() for fitting delay distributions using primarycensored's interval-censored likelihood. Replaces the deprecated estimate_delay().

Changes:

• New estimate_dist() function supporting lognormal, gamma, and weibull distributions
• Vendors primarycensored Stan functions into inst/stan/primarycensored/ for pre-compilation
• Deprecates estimate_delay() with a warning pointing to the new function
• Adds tests for the new function

Related: #1268 (vendored Stan files enable future discretised_pmf work, not included here).

Initial submission checklist

• My PR is based on a package issue and I have explicitly linked it.
• I have tested my changes locally (using devtools::test() and devtools::check()).
• I have added or updated unit tests where necessary.
• I have updated the documentation if required and rebuilt docs if yes (using devtools::document()).
• I have followed the established coding standards (and checked using lintr::lint_package()).
• I have added a news item linked to this PR.

After the initial Pull Request

• I have reviewed Checks for this PR and addressed any issues as far as I am able.

Summary by CodeRabbit

Release Notes

• New Features

  • Added estimate_dist() function for fitting epidemiological delay distributions using MCMC-based Bayesian inference with proper handling of interval censoring and truncation.

• Deprecations

  • Deprecated estimate_delay() in favour of estimate_dist(), which provides improved support for censoring and truncation scenarios.

• Documentation

  • Added comprehensive vignette and documentation for fitting delay distributions with worked examples and interpretation guidance.

[coderabbitai]

Important

Review skipped

Auto incremental reviews are disabled on this repository.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: b87b94cb-0a6d-4674-8003-c032ad562a4e

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

Walkthrough

Adds a new exported function estimate_dist with Stan‑based primary‑censored interval‑censoring support, vendors new Stan functions and a Stan model, updates registration and docs, and deprecates estimate_delay.

Changes

Cohort / File(s)	Summary
Core R implementation R/estimate_dist.R, R/estimate_delay.R	Adds estimate_dist() (implementation, helpers .prepare_delay_intervals, .extract_to_dist_spec, priors, Stan integration); marks estimate_delay() as deprecated and forwards calls with a lifecycle warning.
Stan models & functions inst/stan/estimate_dist.stan, inst/stan/functions/primarycensored.stan	Adds new Stan model estimate_dist.stan and comprehensive primarycensored functions (analytical and numerical likelihoods, truncation handling, ODE fallback, many distribution lcdf/lpdf helpers).
Registration & build NAMESPACE, R/stan.R, R/stanmodels.R	Exports estimate_dist, registers "estimate_dist" in epinow2_stan_model allowed values, and loads the corresponding Stan module.
Documentation & website man/estimate_dist.Rd, man/estimate_delay.Rd, man/dot-*.Rd, man/epinow2_stan_model.Rd, _pkgdown.yml, NEWS.md	Adds user docs for estimate_dist, updates estimate_delay docs to deprecate and point to estimate_dist, documents internal helpers, updates Stan model docs and pkgdown reference, and records changes in NEWS.
Tests tests/testthat/test-estimate_dist.R	Adds tests validating lognormal/gamma parameter recovery, data frame input handling, error cases, max_value behaviour and deprecation warning for estimate_delay.

Possibly related issues

• Vendor primarycensored Stan code for delay distributions #1268 — Vendored primarycensored Stan functions and a new estimate_dist Stan model are added here, matching that issue's intent to vendor and expose primary‑censored functionality.
• Add support for left truncation adjustment in delay tooling (R API only) #1261 — This PR implements truncation/primary‑censoring at the Stan/R boundary and adds left/truncation handling around delay estimation, overlapping the objectives described in the issue.

Possibly related PRs

• Fix flaky integration tests and improve parameter recovery #1252 — Modifies truncation/censoring handling and Stan–R interoperability for delay fitting; closely related to the vendoring and Stan model changes in this PR.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely summarises the main change: adding a new estimate_dist() function with primarycensored support, which is the primary objective of this PR.
Linked Issues check	✅ Passed	The PR successfully addresses all coding objectives from issue #350: implements estimate_dist() as a generalised replacement for estimate_delay(), vendors primarycensored Stan functions, supports lognormal/gamma distributions with proper censoring handling, and includes comprehensive tests.
Out of Scope Changes check	✅ Passed	All changes are within scope. The PR includes the new estimate_dist() function, deprecation of estimate_delay(), Stan vendoring, documentation updates, and tests—all directly aligned with issue #350 objectives. No extraneous changes detected.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch claude/estimate-dist-primarycensored-019U6v1DGjxbeXVpt1eCe7U2

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

Main branch estimate_dist benchmark failed (likely due to interface changes).

[coderabbitai]

Actionable comments posted: 2

🧹 Nitpick comments (3)

tests/testthat/test-estimate_dist.R (1)

264-274: Add an error pattern for consistency.

This test lacks an error message pattern, unlike the other validation tests. Adding a pattern ensures the test catches the correct error rather than any error.

♻️ Proposed fix

   expect_error(
-    estimate_dist(linelist, dist = "weibull")
+    estimate_dist(linelist, dist = "weibull"),
+    "Unsupported distribution"
   )

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@tests/testthat/test-estimate_dist.R` around lines 264 - 274, The test calls
estimate_dist(linelist, dist = "weibull") but uses expect_error without a
pattern; update the assertion to include a regexp that matches the specific
error message thrown by estimate_dist for unsupported distributions (e.g. match
the phrase used in estimate_dist like "unsupported distribution" or the exact
error text it emits) so the test asserts the correct error is raised rather than
any error.

inst/dev/benchmark-estimate-dist.R (1)

21-23: Consider making the benchmark fully reproducible.

The seed list at line 23 uses sample() without a prior set.seed(), so re-running the script produces different seeds each time. This may be intentional for stochastic benchmarking, but if reproducibility across runs is desired, move line 26's set.seed(12345) before line 23.

♻️ Proposed fix for full reproducibility

 ## Configuration
+set.seed(12345)
 n_iter <- 5
 seeds <- sample(.Machine$integer.max, n_iter)
 
 ## Simulate censored delay data
-set.seed(12345)
 n_obs <- 400

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@inst/dev/benchmark-estimate-dist.R` around lines 21 - 23, The script
generates stochastic seeds with seeds <- sample(.Machine$integer.max, n_iter)
but does not set a RNG seed first; to make the benchmark fully reproducible,
call set.seed(12345) (or another chosen constant) before computing seeds (i.e.,
move the existing set.seed(12345) so it runs prior to the seeds <- sample(...)
line) so n_iter and seeds are deterministically generated across runs.

inst/stan/estimate_dist.stan (1)

29-30: Update the dist_id comment to reflect all supported distributions.

The comment states "1=lognormal, 2=gamma" but the R code also supports normal and exponential distributions via primarycensored::pcd_stan_dist_id(). Consider updating the comment to match the actual supported set or simply reference the primarycensored documentation.

📝 Proposed comment update

-  // Distribution ID: 1=lognormal, 2=gamma
+  // Distribution ID: see primarycensored::pcd_stan_dist_id()
+  // Supported: lognormal, gamma, normal, exp
   int<lower=1> dist_id;

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@inst/stan/estimate_dist.stan` around lines 29 - 30, The comment for the
dist_id variable declaration in the Stan file does not document all supported
distributions. Update the comment above the int<lower=1> dist_id parameter to
include all four supported distribution types (lognormal, gamma, normal, and
exponential) with their corresponding ID values, or alternatively reference the
primarycensored::pcd_stan_dist_id() documentation to keep the comment concise
and maintainable.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In @.github/workflows/render-estimate-dist.yaml:
- Around line 10-12: Concurrency group currently uses github.head_ref which is
empty for non-pull_request events; update the concurrency.group expression to
fallback to github.ref when github.head_ref is empty so non-PR triggers have
distinct groups (use the GitHub Actions expression fallback pattern with
github.head_ref || github.ref), keeping cancel-in-progress behavior unchanged.

In `@R/estimate_dist.R`:
- Around line 269-280: The switch handling for init_params in estimate_dist.R
only sets "lognormal" and "gamma", leaving init_params NULL for "normal" and
"exp" and causing pmax/pmin to fail; update the switch(dist, ...) to include MoM
initialisations: for "normal" set init_params to c(mean = wmean, sd =
sqrt(wvar)), and for "exp" set init_params to the exponential parameter using
MoM (rate = 1 / wmean or the parameterization your model uses) so init_params
matches the parameter ordering expected by stan_data; keep the existing clamping
to stan_data$params_lower and stan_data$params_upper and return init_fn <-
function() list(params = init_params) as before.

---

Nitpick comments:
In `@inst/dev/benchmark-estimate-dist.R`:
- Around line 21-23: The script generates stochastic seeds with seeds <-
sample(.Machine$integer.max, n_iter) but does not set a RNG seed first; to make
the benchmark fully reproducible, call set.seed(12345) (or another chosen
constant) before computing seeds (i.e., move the existing set.seed(12345) so it
runs prior to the seeds <- sample(...) line) so n_iter and seeds are
deterministically generated across runs.

In `@inst/stan/estimate_dist.stan`:
- Around line 29-30: The comment for the dist_id variable declaration in the
Stan file does not document all supported distributions. Update the comment
above the int<lower=1> dist_id parameter to include all four supported
distribution types (lognormal, gamma, normal, and exponential) with their
corresponding ID values, or alternatively reference the
primarycensored::pcd_stan_dist_id() documentation to keep the comment concise
and maintainable.

In `@tests/testthat/test-estimate_dist.R`:
- Around line 264-274: The test calls estimate_dist(linelist, dist = "weibull")
but uses expect_error without a pattern; update the assertion to include a
regexp that matches the specific error message thrown by estimate_dist for
unsupported distributions (e.g. match the phrase used in estimate_dist like
"unsupported distribution" or the exact error text it emits) so the test asserts
the correct error is raised rather than any error.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: 00c3eff2-a91f-40cc-9765-00cc3f09c910

📥 Commits

Reviewing files that changed from the base of the PR and between 99cf54d and 80a53a3.

⛔ Files ignored due to path filters (4)

• vignettes/estimate-dist-plot-delays-1.png is excluded by !**/*.png
• vignettes/estimate-dist-prior-predictive-1.png is excluded by !**/*.png
• vignettes/estimate-dist-recovery-density-1.png is excluded by !**/*.png
• vignettes/estimate-dist-recovery-params-1.png is excluded by !**/*.png

📒 Files selected for processing (25)

• .github/workflows/check-primarycensored.yaml
• .github/workflows/render-estimate-dist.yaml
• .github/workflows/stan-model-benchmark.yaml
• NAMESPACE
• NEWS.md
• R/dist_spec.R
• R/estimate_delay.R
• R/estimate_dist.R
• R/get.R
• R/summarise.R
• _pkgdown.yml
• inst/dev/benchmark-estimate-dist.R
• inst/dev/vendor-primarycensored.R
• inst/stan/estimate_dist.stan
• man/dot-get_dist_id.Rd
• man/dot-get_param_names.Rd
• man/dot-prepare_linelist_data.Rd
• man/estimate_delay.Rd
• man/estimate_dist.Rd
• man/get_parameters.Rd
• man/summary.estimate_dist.Rd
• tests/testthat/test-estimate_dist.R
• vignettes/estimate-dist.Rmd
• vignettes/estimate-dist.Rmd.orig
• vignettes/library.bib

✅ Files skipped from review due to trivial changes (10)

• man/dot-get_dist_id.Rd
• man/get_parameters.Rd
• man/dot-get_param_names.Rd
• vignettes/library.bib
• man/summary.estimate_dist.Rd
• man/dot-prepare_linelist_data.Rd
• man/estimate_delay.Rd
• vignettes/estimate-dist.Rmd.orig
• vignettes/estimate-dist.Rmd
• man/estimate_dist.Rd

🚧 Files skipped from review as they are similar to previous changes (3)

• R/estimate_delay.R
• NAMESPACE
• _pkgdown.yml

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 80a53a3 is merged into main:

• ✔️default: 23.7s -> 23.7s [-5.31%, +4.86%]
• ✔️no_delays: 23.7s -> 23.6s [-8.31%, +6.91%]
• ✔️random_walk: 9.25s -> 9.52s [-3.75%, +9.59%]
• ✔️stationary: 14.9s -> 12.4s [-58.12%, +24.36%]
• ✔️uncertain: 36.2s -> 46.2s [-46.5%, +102%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[github-actions]

estimate_infections

Table: Benchmarking results (mean time in seconds). |operation | branch| main| % change| range| trend| |:--------------------|-------:|-------:|--------:|----------:|---------:| |infections | 4.3| 4.4| -1| (-29, 38)| no change| |reports | 4.2| 4.3| -1| (-27, 36)| no change| |delays | 11| 3.5| 207| (122, 353)| slowdown| |report lp | 2.9| 2.9| 0| (-27, 39)| no change| |R0 | 1.3| 1.3| 0| (-28, 36)| no change| |update gp | 1| 1| 0| (-26, 41)| no change| |gt | 0.46| 0.46| 1| (-25, 37)| no change| |day of the week | 0.43| 0.43| 2| (-26, 41)| no change| |truncation | 0.29| 0.29| 1| (-28, 38)| no change| |param lp | 0.24| 0.25| -2| (-28, 37)| no change| |truncate | 0.21| 0.21| 2| (-23, 41)| no change| |delays lp | 0.1| 0.11| -2| (-26, 39)| no change| |rt lp | 0.098| 0.099| 0| (-29, 38)| no change| |gp lp | 0.094| 0.095| 0| (-31, 37)| no change| |generated quantities | 0.0082| 0.0078| 5| (-4, 19)| no change| |assign max | 5.4e-07| 5.8e-07| -6| (-43, 66)| no change|

Main branch estimate_dist benchmark failed (likely due to interface changes).

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 4b82fcf is merged into main:

• ✔️default: 22.6s -> 22.8s [-6.25%, +7.27%]
• ✔️no_delays: 24.2s -> 25s [-15.59%, +22.42%]
• ✔️random_walk: 14.5s -> 9.97s [-79.47%, +16.97%]
• ✔️stationary: 12.4s -> 15.1s [-19.26%, +64.48%]
• ✔️uncertain: 48.4s -> 1.2m [-18.47%, +115.5%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[github-actions]

estimate_infections

Table: Benchmarking results (mean time in seconds). |operation | branch| main| % change| range| trend| |:--------------------|------:|-------:|--------:|----------:|---------:| |infections | 4.1| 4| 7| (-38, 94)| no change| |reports | 4| 3.9| 6| (-38, 87)| no change| |delays | 10| 3.2| 223| (89, 479)| slowdown| |report lp | 3| 2.8| 8| (-37, 93)| no change| |R0 | 1.2| 1.2| 5| (-38, 88)| no change| |update gp | 1| 0.97| 8| (-36, 87)| no change| |day of the week | 0.45| 0.44| 5| (-38, 89)| no change| |gt | 0.46| 0.43| 11| (-37, 106)| no change| |truncation | 0.29| 0.28| 8| (-38, 95)| no change| |param lp | 0.27| 0.25| 12| (-36, 101)| no change| |truncate | 0.21| 0.2| 8| (-39, 88)| no change| |delays lp | 0.13| 0.12| 16| (-32, 115)| no change| |rt lp | 0.1| 0.097| 11| (-39, 90)| no change| |gp lp | 0.088| 0.087| 4| (-42, 85)| no change| |generated quantities | 0.0073| 0.0072| 2| (-7, 9)| no change| |assign max | 6e-07| 6.2e-07| -2| (-45, 45)| no change|

Main branch estimate_dist benchmark failed (likely due to interface changes).

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 300857b is merged into main:

• ✔️default: 23.3s -> 23.8s [-9.28%, +13.64%]
• ✔️no_delays: 26.1s -> 25.8s [-7.51%, +5.47%]
• ✔️random_walk: 13.2s -> 24.2s [-153.42%, +321.08%]
• ✔️stationary: 12.5s -> 12.8s [-19.96%, +24.79%]
• ❗🐌uncertain: 36.6s -> 1.19m [+81.39%, +108.9%]
  Further explanation regarding interpretation and methodology can be found in the documentation.

[github-actions]

estimate_infections

Table: Benchmarking results (mean time in seconds). |operation | branch| main| % change| range| trend| |:--------------------|-------:|-------:|--------:|----------:|---------:| |infections | 4.4| 4.1| 8| (-21, 41)| no change| |reports | 4.3| 4.1| 6| (-21, 36)| no change| |delays | 11| 3.4| 230| (143, 322)| slowdown| |report lp | 3.3| 3| 11| (-19, 58)| no change| |R0 | 1.4| 1.3| 8| (-18, 40)| no change| |update gp | 1.2| 1| 14| (-14, 51)| no change| |gt | 0.51| 0.47| 12| (-16, 47)| no change| |day of the week | 0.47| 0.43| 11| (-17, 61)| no change| |truncation | 0.31| 0.29| 8| (-20, 39)| no change| |param lp | 0.29| 0.27| 11| (-17, 43)| no change| |truncate | 0.24| 0.21| 18| (-8, 54)| no change| |delays lp | 0.12| 0.11| 8| (-19, 44)| no change| |rt lp | 0.12| 0.11| 13| (-20, 67)| no change| |gp lp | 0.11| 0.1| 5| (-26, 39)| no change| |generated quantities | 0.0076| 0.007| 9| (-8, 23)| no change| |assign max | 5.2e-07| 4.9e-07| 10| (-46, 103)| no change|

Main branch estimate_dist benchmark failed (likely due to interface changes).

[seabbs-bot]

Update since last review pass

Bringing reviewers up to speed on what changed when #1356 merged in (plus subsequent lint cleanup on this branch).

Distribution constructors & estimate_dist()

• Exported Exp(), Weibull(), and Normal() constructors.
• estimate_dist() now accepts lognormal / gamma / normal / exp / weibull with default priors and explicit MoM init for each (no more silent NULL fallthrough on normal/exp/weibull).
• init_params wrapped in as.array() so single-parameter (exp) inits become array[1] rather than scalar.
• Init hardened: wvar clamped to sqrt(.Machine$double.eps) for constant delays; non-finite candidates replaced with bounded safe defaults inside Stan-declared bounds. wmean is intentionally not clamped so normal retains negative-mean support.
• dist argument validated before priors so an unsupported distribution surfaces "Unsupported distribution" rather than a priors NULL assertion.

R/dist_spec.R

• Added @importFrom stats pweibull (fixes the Distributions Rd Exp(mean = 4) example failing under as-cran).
• Exp's lower_bounds map gained mean = 0 so Exp(mean = …) validates.
• Dropped the unsupported mean, sd args from Weibull() (the (mean, sd) → (shape, scale) inverse wasn't implemented).
• convert_to_natural if/else-if chain refactored to a switch(), plus assorted indentation fixes.

Verbose gating

• All four cli::cli_inform calls in estimate_dist() (pdate_upr/sdate_upr/obs_date defaults plus the obs_time_threshold Inf-assignment notice) now run under if (verbose).

Vignette / docs

• vignettes/estimate-dist.Rmd.orig: meanlog guidance re-phrased in terms of the median, with the explicit mean correction exp(meanlog + sdlog^2 / 2).
• vignettes/library.bib: bioRxiv → medRxiv for the matching DOI.
• vignettes/estimate-dist.Rmd rendered build no longer leaks "built under R version" warnings (setup chunk now warning = FALSE, message = FALSE).
• man/dot-extract_to_dist_spec.Rd and man/dot-prepare_linelist_data.Rd regenerated with full @param/@return blocks; man/estimate_dist.Rd lists weibull in usage and limitations.
• _pkgdown.yml: registered the delays vignette in the articles index (was missing → pkgdown red).

Tests

• Explicit seed = 12345 on every stan_opts() in tests/testthat/test-estimate_dist.R for deterministic MCMC runs.
• Recovery tests added for normal, exp, weibull.
• correctly handles constant delays without non-finite init gated on skip_integration().
• Snapshots refreshed for simulate-infections and simulate-secondary (the merge had reverted the prior refresh).

Workflows / dev scripts

• .github/workflows/check-primarycensored.yaml: peter-evans/create-pull-request now mints an App token via tibdex/github-app-token@v2 (secrets.APP_ID / APP_PRIVATE_KEY) so the auto-PR triggers downstream CI under the bot identity, matching update-r-multiverse.yaml.
• .github/workflows/stan-model-benchmark.yaml: comment creation gated on the existence of real benchmark output files.
• .github/workflows/render-estimate-dist.yaml: concurrency key now ${{ github.head_ref || github.ref }}.
• inst/dev/benchmark-estimate-dist.R: set.seed() now runs before the sample() it seeds.
• inst/dev/vendor-primarycensored.R: vendor call wrapped in invisible().

NEWS

• Entries added for estimate_dist(), the new Exp/Weibull/Normal constructors, and the estimate_delay() deprecation.

Lint

• Fifteen pre-existing indentation_linter findings in files now in the PR diff (R/checks.R, R/create.R, R/dist_spec.R, R/extract.R, R/fit.R, R/plot.R, R/preprocessing.R, R/simulate_infections.R, R/simulate_secondary.R, R/summarise.R) corrected so lint-changed-files is green.

CI as of 5c76881c — full R-CMD-check matrix (5/5), R-CMD-as-cran-check, test-coverage, pkgdown, codecov all SUCCESS on the previous tip; lint-changed-files queued for this commit.

This is an automated comment summarising bot-merged work; please ping @seabbs for any questions.

[github-actions]

estimate_infections

Table: Benchmarking results (mean time in seconds). |operation | branch| main| % change| range| trend| |:--------------------|-------:|-------:|--------:|---------:|---------:| |infections | 4.1| 4.3| -2| (-53, 34)| no change| |reports | 4| 4.2| -1| (-52, 36)| no change| |delays | 10| 3.5| 205| (49, 321)| slowdown| |report lp | 3.1| 3.1| 3| (-50, 42)| no change| |R0 | 1.3| 1.4| -1| (-50, 40)| no change| |update gp | 1.1| 1.1| 3| (-50, 41)| no change| |gt | 0.47| 0.48| 2| (-50, 42)| no change| |day of the week | 0.44| 0.45| 1| (-50, 41)| no change| |truncation | 0.29| 0.3| -1| (-52, 37)| no change| |param lp | 0.26| 0.28| -2| (-52, 35)| no change| |truncate | 0.21| 0.22| 0| (-53, 37)| no change| |delays lp | 0.12| 0.13| -1| (-53, 40)| no change| |rt lp | 0.12| 0.12| 4| (-52, 50)| no change| |gp lp | 0.099| 0.11| -2| (-51, 55)| no change| |generated quantities | 0.007| 0.0067| 5| (0, 10)| no change| |assign max | 4.7e-07| 4.6e-07| 5| (-44, 78)| no change|

Main branch estimate_dist benchmark failed (likely due to interface changes).

[seabbs]

I think this is ready for review now @sbfnk. Note that we likely need a new issue for adding uncertain pmfs to this vignette and also for allowing convolving delays to retain uncertainty.

[github-actions]

Main branch estimate_dist benchmark failed (likely due to interface changes).

[github-actions]

This is how benchmark results would change (along with a 95% confidence interval in relative change) if 37ade58 is merged into main:

• ✔️default: 23.5s -> 23.6s [-5.94%, +6.44%]
• ✔️no_delays: 26.4s -> 26.2s [-10.91%, +9.96%]
• ✔️random_walk: 27.2s -> 11.3s [-153.04%, +36.15%]
• ✔️stationary: 13.4s -> 13.1s [-12.39%, +7.81%]
• ❗🐌uncertain: 37.4s -> 1.13m [+63.4%, +98.89%]
  Further explanation regarding interpretation and methodology can be found in the documentation.