pub fn betainccinv(a: f64, b: f64, y: f64) -> f64Expand description
Inverse of the complement of the regularized incomplete Beta function, $y = 1 - \I_x(a, b)$
Computes $x \in [0, 1]$ such that 1:
$$ y = 1 - \I_x(a, b) = \I_{1-x}(b, a) , $$
for $0 \le y \le 1$ and $a,b > 0$, with $\I_x(a, b)$ the regularized incomplete Beta function.
This function is the inverse survival function of the Beta distribution.
§Notes
This function uses the incbi Cephes routine 2 for the root of $\I_{1-x}(b, a) = y$,
refined by safeguarded Newton steps, whereas scipy.special.betainccinv wraps the
Boost ibetac_inv routine 3. For example, with $10^{-3} \le a, b \le 10^5$, the
relative error is up to 4e-13, for $a = 10^{-3}$, where the root is ill-conditioned.
Where $1 - \I_x(a, b)$ rounds to $y$ over a range of $x$, e.g. for a subnormal $y$, the result
can be anywhere in that range.
§See also
betaincc: Complement of the regularized incomplete Beta functionbetaincinv: Inverse of the regularized incomplete Beta functionbetainc: Regularized incomplete Beta function $\I_x(a, b)$scipy.special.betainccinv: Corresponding function in SciPy
NIST Digital Library of Mathematical Functions, https://dlmf.nist.gov/8.17.E4. ↩
Cephes Math Library Release 2.4. ↩
The Boost Developers. “Boost C++ Libraries”, https://www.boost.org. ↩