sinistra-core 0.1.0

Drift-diffusion model (DDM) simulator with EZ-diffusion and simulation-based parameter recovery
Documentation

sinistra-core

The simulation and parameter-recovery core of sinistra — a drift-diffusion model (DDM) toolkit for cognitive-science research.

use sinistra_core::{Params, simulate_n, ez_diffusion, fit_simulation, default_fit_guess};

let truth = Params::new(1.2, 1.0, 0.5, 0.25); // drift, boundary, start, non-decision time
let trials = simulate_n(&truth, 100_000, 42); // parallel, deterministic for a given seed

let ez = ez_diffusion(&trials).unwrap();      // closed-form (Wagenmakers et al., 2007)
let sim = fit_simulation(&trials, default_fit_guess(&trials), 200).unwrap(); // simulation-based
  • simulate_trial / simulate_n — Euler–Maruyama diffusion, dt = 1 ms, 10 s cap; simulate_n runs across all cores and is bit-reproducible for a seed (per-trial ChaCha8 cipher streams).
  • ez_diffusion — closed-form inversion of (accuracy, mean RT, RT variance).
  • fit_simulation — Nelder–Mead search that re-simulates the model at each candidate; no closed-form approximation, so no discretization bias.
  • summary_stats — the one (Pc, MRT, VRT) reduction both estimators share.

Full write-up, benchmarks, and a real-data analysis: the repository.

License

MIT.