''' from mle import simulate as sim from statsmodels.graphics.tsaplots import plot_acf, plot_pacf import matplotlib.pyplot as plt import numpy as np from mle import likelihood as logL import scipy.optimize as opt import mle.constraint as cons import datetime np.random.seed(10) phi = [0.1, 0.2, 0.1] x = sim.arpGaussian(t=500, phi=phi) # sigma = 1 x0 = tuple([0. for i in range(len(phi))]) # + [sigma]) bounds = [(-0.99, 0.99) for i in range(len(phi))] # + [(0, None)] # plot_acf(x, lags = 10) # plot_pacf(x, lags = 10) time1 = datetime.datetime.now() # x0 = tuple([0. for i in range(len(phi))]) # bounds = [(-0.99, 0.99) for i in range(len(phi))] # params = opt.minimize(logL.maxARpN, x0 = x0, \ # args = x, \
@author: snake91
'''
import numpy as np
import scipy.linalg as slin
import scipy.optimize as opt
import scipy.stats as st
import pandas as pd
import matplotlib.pyplot as plt
from mle import simulate as sim
from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
from statsmodels.tsa.stattools import acf, pacf
x = sim.arpGaussian(t=500, phi=[0.5, 0.2])
acfList = acf(x, nlags=20)
def empirical_acf(params, n):
rhoList = [1.]
p = len(params)
for i in range(0, n):
rho = 0.
for p in range(len(params)):
rho += params[p] * rhoList[i - p]
'''
Created on Sep 8, 2019
@author: snake91
'''
import pymc3 as pm
import numpy as np
import matplotlib.pyplot as plt
import mle.simulate as sim
np.random.seed(1)
phi1 = 0.1
y = sim.arpGaussian(t=200, phi=[phi1])
def b0_bayesianmodel(y):
with pm.Model(
) as model: # model specifications in PyMC3 are wrapped in a with-statement
# Define priors
# sigma = HalfCauchy('sigma', beta=10, testval=1.)
sigma = 1
intercept = pm.Uniform('phi1', -0.99, 0.99)
x_coeff = phi1 #Normal('x', 0, sd=20)
# Define likelihood
likelihood = pm.Normal('y',
mu=np.hstack([[0], phi1 * y[1:]]),
sd=sigma,
import numpy as np
import matplotlib.pyplot as plt
import scipy.stats as st
import prediction.confidence_interval as ci
window = 250
phi = [-0.6, -0.3]
sigma = 1
u = int(np.random.randint(1000))
# u = 15
np.random.seed(u)
x = np.random.normal(size=500)
np.random.seed(u)
y = sim.arpGaussian(t=500, phi=phi, y0=list(np.random.normal(size=len(phi))))
var_x = [np.percentile(x[i - window:i], q=0.99) for i in range(window, 500)]
var_y_iid = [
np.percentile(y[i - window:i], q=0.99) for i in range(window, 500)
]
var_y_ar = ci.arpGaussian(y, phi=phi, window=250, q=0.99, sigma=1)[1]
# plt.plot(x[window:], label = 'iid')
plt.plot(y[window:], label='arp')
# plt.plot(var_x, label = 'var_iid_iid')
plt.plot(var_y_iid, label='var_arp_iid')
plt.plot(var_y_ar, label='var_arp_arp')
plt.legend()