pypomp.random.binominv

pypomp.random.binominv(u: Array, n: Array, p: Array, exact_max: int = 5, order: int = 2, dtype: dtype | None = None) → Array[source]

Compute the approximate inverse binomial CDF using JAX primitives.

Vectorised implementation using the normal asymptotic expansion formulas from Giles and Beentjes (2024) [1]. A bottom-up exact inverse CDF calculation is performed for small values.

Parameters:

  • u (jax.Array) – Uniform probabilities in [0, 1]. Scalar or array.

  • n (jax.Array) – Number of trials. Must be a positive integer or float.

  • p (jax.Array) – Success probability in [0, 1].

  • exact_max (int, optional) – Maximum iterations for the bottom-up exact inverse CDF stage. Defaults to 5.

  • order (int, optional) – Order of approximation (0, 1, or 2). Defaults to 2.

  • dtype (np.dtype or None, optional) – Floating-point dtype for computation. Inferred from inputs if None.

Returns:

Array of binomial quantiles with the broadcast shape of the inputs.

Return type:

jax.Array

Notes

For speed and accuracy metrics, see the Quant Tests.

See also

fast_binomial

High-level sampler that wraps this function.

References