def tn_pdf(x, loc, scale, lower=0, upper=1):
if upper == np.inf:
Z = 1 - norm._cdf(lower-loc/scale)
else:
Z = norm._cdf((upper-loc)/scale) - norm._cdf((lower-loc)/scale)
pdf_psi = norm._pdf((x - loc) / scale)
return pdf_psi / (scale * Z)
def evaluate(self, t, lamda):
# if not np.all(self._check_args(lamda)):
# raise ValueError('invalid args')
term = np.log((1. - t) / t) * 0.5 / lamda
from scipy.stats import norm
# use special if I want to avoid stats import
transf = ((1 - t) * norm._cdf(lamda + term) +
t * norm._cdf(lamda - term))
return transf
def implied_vol_P(S, K, t, T, q, r, pTrue):
tol = 1e-8
sigmaHat = initial_guess(S, K, t, T, q, r)
sigmaA, sigmaB = sigmaHat, sigmaHat
sigmaDiff = 1.0
n = 1
nMax = 1000
while sigmaDiff >= tol and n < nMax:
if n==1:
p = put_black_scholes(S, K, t, T, sigmaHat, q, r)
else:
p = put_black_scholes(S, K, t, T, sigmaB, q, r)
fn = p-pTrue
fn1 = S*math.exp((0-q)*(T-t))*math.sqrt(T-t)*norm._cdf(dOne(S,K,t,T,sigmaB, q, r))
if fn1 == 0:
return 0.0
increment = fn/fn1 * 0.1
if fn>0:
sigmaA = sigmaB
sigmaB = sigmaB-increment
else:
sigmaMid = (sigmaA+sigmaB)/2
while sigmaA-sigmaB > tol:
sigmaMid = (sigmaA+sigmaB)/2
if put_black_scholes(S, K, t, T, sigmaMid, q, r)-pTrue == 0:
return sigmaMid
elif (put_black_scholes(S, K, t, T, sigmaMid, q, r)-pTrue)*(put_black_scholes(S, K, t, T, sigmaB, q, r)-pTrue)<0:
sigmaA = sigmaMid
else:
sigmaB = sigmaMid
return abs(sigmaMid)
n += 1
sigmaDiff = abs(increment)
return sigmaB
def bias_corrected_ci(true_coeff, samples, conf=95):
"""
Return the bias-corrected bootstrap confidence interval, using the method from the book.
:param true_coeff: The estimates in the original sample
:param samples: The bootstrapped estimates
:param conf: The level of the desired confidence interval
:return: The bias-corrected LLCI and ULCI for the bootstrapped estimates.
"""
ptilde = (samples < true_coeff).mean()
Z = norm.ppf(ptilde)
Zci = z_score(conf)
Zlow, Zhigh = -Zci + 2 * Z, Zci + 2 * Z
plow, phigh = norm._cdf(Zlow), norm._cdf(Zhigh)
llci = np.percentile(samples, plow * 100, interpolation='lower')
ulci = np.percentile(samples, phigh * 100, interpolation='higher')
return llci, ulci
def bias_corrected_ci(estimate: np.array, samples: np.array, conf: float = 95) -> (float, float):
"""
Return the bias-corrected bootstrap confidence interval for an estimate
:param estimate: Numerical estimate in the original sample
:param samples: Nx1 array of bootstrapped estimates
:param conf: Level of the desired confidence interval
:return: Bias-corrected bootstrapped LLCI and ULCI for the estimate.
"""
ptilde = ((samples < estimate) * 1).mean()
Z = norm.ppf(ptilde)
Zci = z_score(conf)
Zlow, Zhigh = -Zci + 2 * Z, Zci + 2 * Z
plow, phigh = norm._cdf(Zlow), norm._cdf(Zhigh)
llci = np.percentile(samples, plow * 100, interpolation='lower')
ulci = np.percentile(samples, phigh * 100, interpolation='higher')
return llci, ulci
def bias_corrected_ci(estimate, samples, conf=95):
"""
Return the bias-corrected bootstrap confidence interval for an estimate
:param estimate: Numerical estimate in the original sample
:param samples: Nx1 array of bootstrapped estimates
:param conf: Level of the desired confidence interval
:return: Bias-corrected bootstrapped LLCI and ULCI for the estimate.
"""
# noinspection PyUnresolvedReferences
ptilde = ((samples < estimate) * 1).mean()
Z = norm.ppf(ptilde)
Zci = z_score(conf)
Zlow, Zhigh = -Zci + 2 * Z, Zci + 2 * Z
plow, phigh = norm._cdf(Zlow), norm._cdf(Zhigh)
llci = np.percentile(samples, plow * 100, interpolation="lower")
ulci = np.percentile(samples, phigh * 100, interpolation="higher")
return llci, ulci
def transform_hr(t, lamda):
'''model of Huesler Reiss 1989
special case: a1=a2=1 : symmetric negative logistic of Galambos 1978
restrictions:
- lambda in (0,inf)
'''
def _check_args(lamda):
cond = (lamda > 0)
return cond
if not np.all(_check_args(lamda)):
raise ValueError('invalid args')
term = np.log((1. - t) / t) * 0.5 / lamda
from scipy.stats import norm #use special if I want to avoid stats import
transf = (1 - t) * norm._cdf(lamda + term) + t * norm._cdf(lamda - term)
return transf
def transform_hr(t, lamda):
'''model of Huesler Reiss 1989
special case: a1=a2=1 : symmetric negative logistic of Galambos 1978
restrictions:
- lambda in (0,inf)
'''
def _check_args(lamda):
cond = (lamda > 0)
return cond
if not np.all(_check_args(lamda)):
raise ValueError('invalid args')
term = np.log((1. - t) / t) * 0.5 / lamda
from scipy.stats import norm #use special if I want to avoid stats import
transf = (1 - t) * norm._cdf(lamda + term) + t * norm._cdf(lamda - term)
return transf
def estimate_power(self):
"""
Estimate the power of the tests entered in the-pcurve, correcting for publication bias.
:return: p, lbp, ubp: The power, and (lower bound, upper bound) of the 95% CI for power.
"""
p = self._solve_power_for_pct(.50)
p05 = norm._cdf(self._compute_stouffer_z_at_power(.051))
p99 = norm._cdf(self._compute_stouffer_z_at_power(.99))
if p05 <= .95:
lbp = .05
elif p99 >= .95:
lbp = .99
else:
lbp = self._solve_power_for_pct(.95)
if p05 <= .05:
ubp = .05
elif p99 >= .05:
ubp = .99
else:
ubp = self._solve_power_for_pct(.05)
return p, (lbp, ubp)
def implied_vol_C(S, K, t, T, q, r, cTrue):
tol = 1e-8
sigma = initial_guess(S, K, t, T, q, r)
sigmaDiff = 1.0
n = 1
nMax = 1000
while sigmaDiff >= tol and n < nMax :
#f(xn)
c = call_black_scholes(S, K, t, T, sigma, q, r)
fn = c - cTrue
#f'(xn)
fn1 = S * math.exp((0 - q) * (T - t)) * math.sqrt(T - t) * norm._cdf(dOne(S,K,t,T,sigma, q, r))
increment = fn / fn1 * 0.1
sigma = sigma - increment
n += 1
sigmaDiff = abs(increment)
return sigma
def gaussian_cdf(h, Xi, x):
return h * norm._cdf((x - Xi)/h)
def Increment(option_price,true_data,Stock,d1):
increment = (option_price-true_data)/(Stock*math.exp(-q*(maturity-float(t)))*math.sqrt(maturity-float(t)*norm._cdf(d1)))
return increment
def Increment(option_price, true_data, Stock, d1):
increment = (option_price -
true_data) / (Stock * math.exp(-q * (maturity - float(t))) *
math.sqrt(maturity - float(t) * norm._cdf(d1)))
return increment
rate=0.01
repo = 0
tol = 1e-8
i=0
sigmadiff=1
sigmahat = math.sqrt(2*abs((math.log(stock/strike)+rate*maturity)/maturity))
sigma = sigmahat
#print sigmahat
while sigmadiff>=tol and i<=100:
#d1= ((math.log(stock/strike)+(rate-repo)*(maturity-float(t)))/sigma*math.sqrt(maturity-float(t)))+0.5*sigma*math.sqrt(maturity-float(t))
d1 = (math.log(stock/strike)+float(rate-repo)*maturity)/(float(sigma)*math.sqrt(maturity))+0.5*float(sigma)*math.sqrt(maturity)
#d2= ((math.log(stock/strike)+(rate-repo)*(maturity-float(t)))/sigma*math.sqrt(maturity-float(t)))-0.5*sigma*math.sqrt(maturity-float(t))
d2 = (math.log(stock/strike)+float(rate-repo)*maturity)/(float(sigma)*math.sqrt(maturity))-0.5*float(sigma)*math.sqrt(maturity)
call_price = call_option_price(stock, strike, maturity, rate, repo)
increment = (call_price-c_true)/(stock*math.exp(-repo*(maturity-float(t)))*math.sqrt(maturity-float(t)*norm._cdf(d1)))
sigma = sigma-increment
sigmadiff=abs(increment)
i=i+1
print "---------------------------------"
print 'd1: %.8f'% d1
print 'd2: %.8f'% d2
print 'Call price: %.8f'% call_price
print 'sigma: %.8f'% sigma
def mytruncpdf_01bounds(x, loc, scale):
Z = norm._cdf((1-loc)/scale) - norm._cdf(-loc/scale)
pdf_psi = norm._pdf((x - loc) / scale)
return pdf_psi / (scale * Z)
def my_tn_cdf(upper, mn, scale):
"""Assumes that clipping between 0 and 1.
Some notation from wiki page on truncated normal"""
norm_cdf_alpha = norm._cdf(-mn/scale)
return ((norm._cdf((upper - mn) / scale) - norm_cdf_alpha) /
(norm._cdf((1 - mn) / scale) - norm_cdf_alpha))
def tn_pdf_01(x, loc, scale):
Z = norm._cdf((1-loc)/scale) - norm._cdf(-loc/scale)
pdf_psi = norm._pdf((x - loc) / scale)
return pdf_psi / (scale * Z)
