def test_0():
out = fs.static(trial=trial, trials=500)
means = out.mean(dim='trials')
meanEps = float(means.loc['eps'])
meanB = float(means.loc['b'])
assert abs(meanEps - 13.0) < 13.0
assert abs(meanB - 26.0) < 0.01
def test_normal_0():
# specify correlation matrix
rho = np.array([[1.0, 0.5], [0.5, 1.0]])
# set up function to perform a single trial
def f(draw):
eps = fs.normal(rho, draw) # vector of two correlated stand. normal draws
return {"eps0": eps[0], "eps1": eps[1]}
# perform simulations
out = fs.static(trial=f, trials=2000)
sampcorr = np.corrcoef(out, rowvar=False)[0, 1]
assert abs(sampcorr - 0.5) < 0.05
sys.path.append(
os.path.join(os.path.dirname(os.path.realpath(__file__)), "../"))
import numpy as np
import funcsim as fs
def trial(draw):
# function to perform one trial.
# simulate one std. norm variable, and one Bernoilli variable
# indep. uniform draw using built-in stratified/Latin hypercube sampling
u_LHS = next(draw)
# indep. uniform draw using non-stratified sampling
u_no_LHS = np.random.uniform(0.0, 1.0)
# return dict with var names and values
return {"u_LHS": u_LHS, "u_no_LHS": u_no_LHS}
# simulate
out = fs.static(trial=trial, trials=100)
# sort simulated u values, send to screen
srtd_u_LHS = np.sort(out[:, 0])
srtd_u_no_LHS = np.sort(out[:, 1])
print("\nu_LHS, u_no_LHS")
for i in range(len(srtd_u_LHS)):
print("%s, %s" % (srtd_u_LHS[i], srtd_u_no_LHS[i]))
import funcsim as fs
import numpy as np
# https://en.wikipedia.org/wiki/Monte_Carlo_integration
def h(x, y):
return 1.0 if x**2.0 + y**2.0 < 1.0 else 0.0
def trial(draw):
x = 2.0 * next(draw) - 1.0 # uniform over [-1, 1]
y = 2.0 * next(draw) - 1.0
return {"h": h(x, y)}
"""
out = fs.crosssec(trial=trial, trials=5000, multi=False)
area = float(out.mean()) # area of a quarter circle
sigf = float(out.std())
n = len(out)
print("n: %s" % n)
print("value of pi/4 is approximately %s" % area)
print("sig_f: %s" % sigf)
print("sig_area: %s" % (sigf / n**0.5))
"""
for s in range(100):
out = fs.static(trial=trial, trials=5000, multi=False, seed=s)
area = float(out.mean()) # area of a quarter circle
print(area)
os.path.join(os.path.dirname(os.path.realpath(__file__)), "../"))
from scipy import stats
import funcsim as fs
def trial(draw, prob):
# function to perform one trial.
# simulate one std. norm variable, and one Bernoilli variable that is 1
# with probability 'prob'
# independent uniform draws
u1 = next(draw)
u2 = next(draw)
# inverse CDF transformations
eps = stats.norm.ppf(u1)
b = stats.bernoulli.ppf(u2, prob)
# return dict with var names and values
return {"eps": eps, "b": b}
# simulate for first scenario, with 'prob' = 0.25
out = fs.static(trial=lambda dr: trial(dr, 0.25), trials=15, multi=True)
print("\nMeans of the simulated variables:\n%s" % out.mean(dim='trials'))
# simulate for second scenario, with 'prob' = 0.5
out = fs.static(trial=lambda dr: trial(dr, 0.5), trials=15, multi=True)
print("\nMeans of the simulated variables:\n%s" % out.mean(dim='trials'))