Automatic Differentiation
 
Loading...
Searching...
No Matches
inv_inc_beta.hpp
Go to the documentation of this file.
1#ifndef STAN_MATH_FWD_FUN_INV_INC_BETA_HPP
2#define STAN_MATH_FWD_FUN_INV_INC_BETA_HPP
3
18
19namespace stan {
20namespace math {
21
35template <typename T1, typename T2, typename T3,
36 require_all_stan_scalar_t<T1, T2, T3>* = nullptr,
37 require_any_fvar_t<T1, T2, T3>* = nullptr>
39 const T2& b,
40 const T3& p) {
41 using T_return = partials_return_t<T1, T2, T3>;
42 auto a_val = value_of(a);
43 auto b_val = value_of(b);
44 auto p_val = value_of(p);
45 T_return w = inv_inc_beta(a_val, b_val, p_val);
46 T_return log_w = log(w);
47 T_return log1m_w = log1m(w);
48 auto one_m_a = 1 - a_val;
49 auto one_m_b = 1 - b_val;
50 T_return one_m_w = 1 - w;
51 auto ap1 = a_val + 1;
52 auto bp1 = b_val + 1;
53 auto lbeta_ab = lbeta(a_val, b_val);
54 auto digamma_apb = digamma(a_val + b_val);
55
56 T_return inv_d_(0);
57
59 std::vector<T_return> da_a{a_val, a_val, one_m_b};
60 std::vector<T_return> da_b{ap1, ap1};
61 auto da1 = exp(one_m_b * log1m_w + one_m_a * log_w);
62 auto da2 = exp(a_val * log_w + 2 * lgamma(a_val)
63 + log(hypergeometric_3F2(da_a, da_b, w)) - 2 * lgamma(ap1));
64 auto da3 = inc_beta(a_val, b_val, w) * exp(lbeta_ab)
65 * (log_w - digamma(a_val) + digamma_apb);
66 inv_d_ += forward_as<fvar<T_return>>(a).d_ * da1 * (da2 - da3);
67 }
68
70 std::vector<T_return> db_a{b_val, b_val, one_m_a};
71 std::vector<T_return> db_b{bp1, bp1};
72 auto db1 = (w - 1) * exp(-b_val * log1m_w + one_m_a * log_w);
73 auto db2 = 2 * lgamma(b_val) + log(hypergeometric_3F2(db_a, db_b, one_m_w))
74 - 2 * lgamma(bp1) + b_val * log1m_w;
75
76 auto db3 = inc_beta(b_val, a_val, one_m_w) * exp(lbeta_ab)
77 * (log1m_w - digamma(b_val) + digamma_apb);
78
79 inv_d_ += forward_as<fvar<T_return>>(b).d_ * db1 * (exp(db2) - db3);
80 }
81
83 inv_d_ += forward_as<fvar<T_return>>(p).d_
84 * exp(one_m_b * log1m_w + one_m_a * log_w + lbeta_ab);
85 }
86
87 return fvar<T_return>(w, inv_d_);
88}
89
90} // namespace math
91} // namespace stan
92#endif
auto hypergeometric_3F2(const Ta &a, const Tb &b, const Tz &z)
Hypergeometric function (3F2).
fvar< T > lbeta(const fvar< T > &x1, const fvar< T > &x2)
Definition lbeta.hpp:14
T value_of(const fvar< T > &v)
Return the value of the specified variable.
Definition value_of.hpp:18
fvar< T > log(const fvar< T > &x)
Definition log.hpp:18
fvar< partials_return_t< T1, T2, T3 > > inv_inc_beta(const T1 &a, const T2 &b, const T3 &p)
The inverse of the normalized incomplete beta function of a, b, with probability p.
fvar< T > inc_beta(const fvar< T > &a, const fvar< T > &b, const fvar< T > &x)
Definition inc_beta.hpp:19
fvar< T > lgamma(const fvar< T > &x)
Return the natural logarithm of the gamma function applied to the specified argument.
Definition lgamma.hpp:21
fvar< T > log1m(const fvar< T > &x)
Definition log1m.hpp:12
fvar< T > digamma(const fvar< T > &x)
Return the derivative of the log gamma function at the specified argument.
Definition digamma.hpp:23
fvar< T > exp(const fvar< T > &x)
Definition exp.hpp:15
typename partials_return_type< Args... >::type partials_return_t
The lgamma implementation in stan-math is based on either the reentrant safe lgamma_r implementation ...
Defines a static member function type which is defined to be false as the primitive scalar types cann...
Definition is_fvar.hpp:15
This template class represents scalars used in forward-mode automatic differentiation,...
Definition fvar.hpp:40