SSRTcalc 2.1.1

Data

• Refined the fixed dataset: excluded subject 2 (no recorded responses on any trial) and one invalid stop trial (recorded reaction time of 0 ms), leaving 50 subjects and 28,799 trials. Go-trial soa is now NA, matching the adaptive dataset convention. See ?fixed.

New features

Monte Carlo extensions (R/mc_extensions.R)

• ssrt_boot() – nonparametric bootstrap confidence intervals for any SSRT estimator (integration_* or mean_*), with print() and plot() methods.
• ssrt_simulate() – parametric Monte Carlo via ex-Gaussian simulation of the horse-race model, with optional parameter-recovery checking via ssrt_true.
• ssrt_power() – minimum-trial-count / power analysis: simulates datasets across a range of stop-trial counts and reports SE and RMSE of the SSRT estimate as a function of sample size.
• ssrt_robustness() – sensitivity analysis for three violations of the independent horse-race assumptions: go/stop process correlation, trigger failures, and go-RT slowing on stop trials (“stop-signal cost”).
• run_all_mc() – convenience wrapper running all four analyses above.

Bayesian / Stan extensions (R/stan_extensions.R, inst/stan/)

• ssrt_stan() – fits the independent horse-race model with ex-Gaussian go- and stop-process distributions via Hamiltonian Monte Carlo.
  • model = "single" (default) or model = "hierarchical" (non-centred parameterisation across subjects).
  • trigger_failure = TRUE adds the trigger-failure parameter of Matzke et al. (2013).
  • Works with either cmdstanr (recommended) or rstan as backend.
• print(), summary(), plot(), coef() methods for ssrt_stan objects, including posterior densities, trace plots, pairs plots (single-subject) and a forest plot of per-subject SSRT (hierarchical).
• ranef() – new S3 generic (SSRTcalc previously had no dependency that provided this), with a method for hierarchical ssrt_stan fits returning per-subject posterior summaries.
• ssrt_stan_pp_check() – posterior predictive check plots for go RT and SSRT.
• ssrt_stan_inhibition_fn() – posterior inhibition function P(inhibit | SSD) with a 90% credible band.
• ssrt_stan_compare() – fits the base and trigger-failure models side by side and tabulates posterior SSRT summaries.
• ssrt_stan_loo() – LOO-CV via the loo package (requires log_lik in generated quantities; the function explains how to add it if absent).
• Two bundled Stan programs in inst/stan/: ssrt_single.stan and ssrt_hier.stan.

Internals

• R/helpers.R – shared internal utilities: ex-Gaussian fitting (.fit_exgaussian), sampling (.rexgaussian), CDF (.pexgaussian), trial splitting (.split_trials), numerical inhibition-probability quadrature (.p_inhibit_r), and posterior-draw extraction helpers shared between the MC and Stan extensions.
• R/data.R – documentation for the bundled adaptive dataset.

Other changes

• Added a full testthat (edition 3) test suite covering the core estimators, all four Monte Carlo functions, and the pure-R Stan helpers (end-to-end Stan fitting tests are skipped automatically if neither cmdstanr nor rstan is installed).
• Imports now includes graphics and grDevices (used by the new plot.* methods).
• Fixed package metadata (NAMESPACE imports, ORCID).

SSRTcalc 0.3.3

• Initial CRAN release: integration_adaptiveSSD(), integration_fixedSSD(), mean_adaptiveSSD(), mean_fixedSSD().