def test_3(): # many steps (check that recursion does not bust stack)
out = fs.recdyn(step=step,
data0=data0(),
steps=2000,
trials=10,
stdnorm=True)
assert type(out) == xr.DataArray
def test_2(): # alternative seed
out = fs.recdyn(step=step,
data0=data0(),
steps=10,
trials=500,
seed=123,
stdnorm=True)
assert type(out) == xr.DataArray
assert abs(float(out[:, 0, 10].mean()) - 1.0234) < 0.01
def test_1(): # use multi
out = fs.recdyn(step=step,
data0=data0(),
steps=10,
trials=500,
multi=True,
stdnorm=True)
assert type(out) == xr.DataArray
assert abs(float(out[:, 0, 10].mean()) - 1.0234) < 0.01
def test_0(): # basic
out = fs.recdyn(step=step,
data0=data0(),
steps=10,
trials=500,
stdnorm=True)
assert type(out) == xr.DataArray
print(out)
print(out[:, 0, 10].mean())
assert abs(float(out[:, 0, 10].mean()) - 1.0234) < 0.01
def step(data, draw):
# take one step through time
# value of p in previous period
pLag1 = fs.recall(data, "p", lag=1)
# uniform draw --> standard normal draw
u = next(draw)
eps = stats.norm.ppf(u)
# update all intermediate variables
pNew = gbm(s0=pLag1, dt=1.0 / 12.0, mu=0.01, sig=0.10, eps=eps)
cNew = max(0.0, pNew - 1.0)
# return updated price history
dataNew = fs.chron(data, {"p": pNew, "c": cNew})
return dataNew
# set up existing/historical data
steps = [0, 1, 2]
variables = ["p", "c"]
a = np.array([[1.0, np.nan], [1.01, np.nan], [0.99, np.nan]])
data0 = \
xr.DataArray(data=a, coords=(('steps', steps), ('variables', variables)))
# simulate
out = fs.recdyn(step=step, data0=data0, steps=10, trials=20, multi=True)
print(out)