Changelog • DrWrinch

DrWrinch 0.0.1.9001

Added sens_coding() – the coding-error sensitivity tipping point for both models: the smallest number of pro-working-theory observations that would have to be re-coded as pro-rival before the Bayes factor drops below the threshold. This is the paper’s third sensitivity question, alongside observation bias (omega_star) and the rival-tilted prior (M_star). It is also surfaced in the Shiny app’s Sensitivity tab.
Refreshed the running example throughout (function examples, vignette, README, Shiny defaults) to the paper’s (y_W = 9, y_R = 3) case, matching the submitted manuscript: the binomial Bayes factor is 20.67 (just above the threshold of 20) and the hypergeometric is 323. The earlier (7, 3) example predated a revision of the paper.
inst/CITATION now points to the preprint, arXiv:2606.16683.

Added run_app() – an interactive Shiny app exposing the four core functions through a browser UI. Inputs (y_W, y_R, threshold) sit in a sidebar with theta_cut, prior_a, and prior_b tucked into an Advanced accordion. The main panel has three tabs: Result (Binomial / Urn sub-tabs with formatted Bayes factors, Kass & Raftery verdicts, and direction labels), Sensitivity (tipping- point prose plus plotly curves of BF vs. omega and BF vs. M), and About (running example, verdict scale, and sources).
Added shiny (>= 1.7.0), bslib (>= 0.6.0), and plotly (>= 4.10.0) to Suggests. The Shiny stack is runtime-only; the four core Bayes factor functions remain usable without it. run_app() guards each dependency with requireNamespace() and reports a clear install hint on miss.
Dropped the unused DrBristol cross-reference from Suggests (it was never library()-ed in any R, test, or vignette code and was blocking the GitHub Actions workflows).
Repository published at https://github.com/bowers-illinois-edu/DrWrinch; pkgdown site at https://bowers-illinois-edu.github.io/DrWrinch/; Shiny app deployed to https://jakebowers.shinyapps.io/drwrinch/.

CRAN-readiness additions: Depends: R (>= 4.0.0), cph role on all authors, inst/CITATION pointing to the working paper.
Added vignettes/getting-started.Rmd walking through the paper running example with both models and the sensitivity tipping points.
Added cran-comments.md for submission.
Added GitHub Actions workflows for R CMD check (matrix across Ubuntu / macOS / Windows) and pkgdown deployment.
Added _pkgdown.yml for the package website.

Implemented bf_urn() (Formulation C of the hypergeometric urn model: WTF = (y_W + 1, max(1, y_R)), RTF = (y_W, y_W + 1)).
Extended bf_binomial() with observation-bias parameter omega and general Beta priors.
Implemented sens_binomial() and sens_urn() (bias and prior tipping points via uniroot).
Added tests/testthat/ covering the substantive properties: BF = 1 under symmetric evidence, the paper’s analytical value bf_urn(7, 3) = 39, RTF-degeneracy at y_R > y_W + 1, pseudo-observation prior equivalence, and parity with the paper’s sens_bias_hyper_separated.