15e4175f1bf81eb7af4f5c5086100460  a35_u9_check.py
S M_ode: True
S M0: True
S stein_pos: True
S f_formula: True
S Eh_zero: True
S odd: True
S f2: True
S f3: True
S f1_equals_h_plus_wf: True
S B_identity: True
S B_f2_gap: True
S B_f3_sign: True
S B_limit: True
PART S PASS
A c*M(0) - 1 = [+/- 2.44e-38]
A sample points in (0, 60]: 1040, not certified: 0
A f''(1e-12) = [0.9999999999984042308783957692882402147 +/- 7.60e-38]
A C5 = [0.8551004554708192457185191123470460691 +/- 3.87e-38]  2(C5 - c) = [0.1144317893359077796772539849565646642 +/- 8.23e-38]
A 2(sqrt(2/e) - c) = [0.1197586483156828812005950428200174823 +/- 6.20e-38] (RQ-B Prop. 6(b) threshold; 0.11976 exceeds it: True )
PART A PASS
E two-sided inequality: 3000 random K, violations 0, ratio range [-0.3166, 0.0861]
E zero-bias identity E[Z f(Z)] = E f'(Z + U X), 300 compound Poisson K: max |lhs - rhs| = 3.05e-15
E zero-bias identity with Gaussian part 0.6 (adaptive quadrature split at the kinks): lhs -0.393993919462 rhs -0.393993919462
E sup |f'(a) - f'(b)|/|a - b| over 200,000 pairs = 0.997487; Stein equation residual on [-8, 8] = 8.88e-16
PART E PASS
H sign +1: max of sign*(E|Z| - c)/((1/2) int|x| dK) = 0.08613 at atoms [0.12, -1.368], weights [np.float64(0.0), np.float64(1.0)], gaussian 0.0000
H sign -1: max of sign*(E|Z| - c)/((1/2) int|x| dK) = 0.35337 at atoms [-1.9692, 1.9692], weights [np.float64(0.5), np.float64(0.5)], gaussian 0.0000
PART H: NO VIOLATION

real	0m12.661s
user	0m13.621s
sys	0m0.228s
