Issues

Fused C++ adjoint for primarycensored_sone_pmf_vectorized()

epinowcast/primarycensored#426 · opened

on Nov 2, 2026

Use the C++ uniform lcdf in the scalar primarycensored_lcdf() path

epinowcast/primarycensored#425 · opened

on Nov 2, 2026

Review guide for the open PR stack

epinowcast/primarycensored#420 · opened

on Oct 26, 2026

pcens_cdf.default() errors with a narrow truncated logistic primary

epinowcast/primarycensored#416 · opened

on Oct 1, 2026

Analytical CDF returns NaN for the Prentice generalised gamma with small Q

epinowcast/primarycensored#415 · opened

on Oct 1, 2026

Apply the delay argument guard to the existing analytical pcens_cdf() methods

epinowcast/primarycensored#412 · opened

on Sep 30, 2026

Lognormal uniform terms drop one term alone near the log CDF underflow point

epinowcast/primarycensored#411 · opened

on Sep 30, 2026

Stan numerical path and R gamma analytic path lose accuracy in the lower tail when d < pwindow

epinowcast/primarycensored#406 · opened

on Sep 30, 2026

1

Upper tail gradient of the log PMF is wrong where the log CDF is close to 0

epinowcast/primarycensored#404 · opened

on Sep 30, 2026

NaN log PMF from upper-tail rounding in the analytical log CDF for large shapes

epinowcast/primarycensored#401 · opened

on Sep 30, 2026

Bounded transform for exponential and gamma delays with tilts above the rate

epinowcast/primarycensored#399 · opened

on Sep 30, 2026

Stan gamma uniform primary log CDF returns NaN in the deep lower tail for large shapes

epinowcast/primarycensored#394 · opened

on Sep 30, 2026

Stan uniform primary gradients with respect to shape are inaccurate for small Weibull and generalised gamma shapes

epinowcast/primarycensored#393 · opened

on Sep 30, 2026

Stan uniform primary log CDFs are wrong in the far upper tail

epinowcast/primarycensored#392 · opened

on Sep 30, 2026

Numerical integration over [0, pwindow] is inaccurate for non-uniform primary when x < pwindow

epinowcast/primarycensored#389 · opened

on Sep 30, 2026

Stable closed forms for inadmissible tilts of the exponentially tilted primary

epinowcast/primarycensored#388 · opened

on Sep 30, 2026

Cancellation error in the uniform terms difference for large k * shape

epinowcast/primarycensored#385 · opened

on Sep 30, 2026

tests: Stan analytical generalised gamma test compares the analytical path with itself

epinowcast/primarycensored#384 · opened

on Sep 30, 2026

update.pcens() accepts gamma rate and scale together on the analytical path

epinowcast/primarycensored#382 · opened

on Sep 30, 2026

Gamma delay lower-tail underflow and gamma_lcdf gradient failures for large shape

epinowcast/primarycensored#381 · opened

on Sep 30, 2026

1

Reduce per-call cost of the existing uniform primary analytic solutions

epinowcast/primarycensored#378 · opened

on Sep 30, 2026

1

Export a reusable grouped log-likelihood helper for use in external optimisers

epinowcast/primarycensored#376 · opened

on Sep 30, 2026

1

Document the transform representation and tractability table in the analytic solutions vignette

epinowcast/primarycensored#375 · opened

on Sep 30, 2026

Transform-based solutions for Erlang, exponential mixture and phase-type delays

epinowcast/primarycensored#374 · opened

on Sep 30, 2026

Transform-based solutions for delay families without Stan built-ins (inverse Gaussian, Gompertz, Burr)

epinowcast/primarycensored#373 · opened

on Sep 30, 2026

Navigation Menu

Repository navigation

Issues

Fused C++ adjoint for primarycensored_sone_pmf_vectorized()

Use the C++ uniform lcdf in the scalar primarycensored_lcdf() path

Review guide for the open PR stack

pcens_cdf.default() errors with a narrow truncated logistic primary

Analytical CDF returns NaN for the Prentice generalised gamma with small Q

Apply the delay argument guard to the existing analytical pcens_cdf() methods

Lognormal uniform terms drop one term alone near the log CDF underflow point

Stan numerical path and R gamma analytic path lose accuracy in the lower tail when d < pwindow

Upper tail gradient of the log PMF is wrong where the log CDF is close to 0

NaN log PMF from upper-tail rounding in the analytical log CDF for large shapes

Bounded transform for exponential and gamma delays with tilts above the rate

Stan gamma uniform primary log CDF returns NaN in the deep lower tail for large shapes

Stan uniform primary gradients with respect to shape are inaccurate for small Weibull and generalised gamma shapes

Stan uniform primary log CDFs are wrong in the far upper tail

Numerical integration over [0, pwindow] is inaccurate for non-uniform primary when x < pwindow

Stable closed forms for inadmissible tilts of the exponentially tilted primary

Cancellation error in the uniform terms difference for large k * shape

tests: Stan analytical generalised gamma test compares the analytical path with itself

update.pcens() accepts gamma rate and scale together on the analytical path

Gamma delay lower-tail underflow and gamma_lcdf gradient failures for large shape

Reduce per-call cost of the existing uniform primary analytic solutions

Export a reusable grouped log-likelihood helper for use in external optimisers

Document the transform representation and tractability table in the analytic solutions vignette

Transform-based solutions for Erlang, exponential mixture and phase-type delays

Transform-based solutions for delay families without Stan built-ins (inverse Gaussian, Gompertz, Burr)