def test_sqdistalgorithm_additive1():
"""SqDistAlgorithm_test.test_sqdistalgorithm_additive1()
Uses a simple 12 point data series to compare additive inputs with
corresponding outputs.
"""
# configure to test zero-step predictions of 4 "season" cycles
m = 4
t = np.linspace(0, 2 * np.pi, m + 1)[:-1]
hstep = 0
# initial slope is 0; average age is infinite
b0 = 0
beta = 1 / np.inf
# initial trendline is 0; average age is 12 steps
l0 = 0
alpha = 1 / 12.0
# initial seasonal correction is sinusoid; average age is 12 steps
s0 = np.sin(t)[0:4]
gamma = 1 / 12.0 * m
# standard deviation of unit-amplitude sinusoid
sigma0 = [np.sqrt(0.5)]
# predict three cycles ahead given l0 and s0, no inputs,
# and assume PI only grows with trendline adjustments
yobs1 = np.zeros(12) * np.nan
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(yobs1,
m,
alpha=alpha,
beta=beta,
gamma=0,
s0=s0,
l0=l0,
b0=b0,
sigma0=sigma0,
hstep=hstep)
assert_almost_equal(yhat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
err_msg='yhat1 should almost equal simple time series')
assert_almost_equal(shat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
err_msg='shat1 should almost equal simple time series')
assert_almost_equal(
sighat1, [
0.70710678, 0.70955777, 0.71200031, 0.71443451, 0.71686044,
0.71927819, 0.72168784, 0.72408947, 0.72648316, 0.72886899,
0.73124703, 0.73361737
],
err_msg='sighat1 should almost equal simple time series')
# predict three cycles ahead given l0 and s0, no inputs,
# and assume PI only grows with seasonal adjustments
yobs1 = np.zeros(12) * np.nan
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(yobs1,
m,
alpha=0,
beta=0,
gamma=gamma,
s0=s0,
l0=0,
b0=0,
sigma0=sigma0,
hstep=hstep)
assert_almost_equal(
yhat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
err_msg='yhat1 should almost equal simple time series, 2nd run')
assert_almost_equal(
shat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
err_msg='shat1 should almost equal simple time series, 2nd run')
assert_almost_equal(
sighat1, [
0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.74535599,
0.74535599, 0.74535599, 0.74535599, 0.78173596, 0.78173596,
0.78173596, 0.78173596
],
err_msg='sighat1 should almost equal simple time series, 2nd run')
# smooth three cycles' worth of zero-value input observations,
# assuming only the trendline varies
yobs1 = np.zeros(12)
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(yobs1,
m,
alpha=alpha,
beta=0,
gamma=0,
s0=s0,
l0=0,
b0=0,
sigma0=sigma0,
hstep=hstep)
# check output
assert_almost_equal(
yhat1, [
0, 1, -0.08333333, -1.07638889, 0.01331019, 1.01220100,
-0.07214908, -1.06613666, 0.02270806, 1.02081573, -0.06425225,
-1.0588979
],
8,
err_msg='yhat1 should almost equal simple time series, 3rd run')
assert_almost_equal(
shat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
8,
err_msg='shat1 should almost equal simple time series, 3rd run')
assert_almost_equal(
sighat1, [
0.64818122, 0.67749945, 0.62798561, 0.66535255, 0.61101568,
0.64444779, 0.59675623, 0.63587127, 0.58477433, 0.62111112,
0.57470621, 0.61505552
],
8,
err_msg='sighat1 should almost equal simple time series, 3rd run')
# smooth three cycles' worth of zero-value input observations,
# assuming only the seasonal adjustments vary
yobs1 = np.zeros(12)
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(yobs1,
m,
alpha=0,
beta=0,
gamma=gamma,
s0=s0,
l0=0,
b0=0,
sigma0=sigma0,
hstep=hstep)
# check output
assert_almost_equal(
yhat1, [
0, 1, 0, -1, 0, 0.66666667, 0, -0.66666667, 0, 0.44444444, 0,
-0.44444444
],
8,
err_msg='yhat1 should almost equal simple time series, 4th run')
assert_almost_equal(
shat1, [
0, 1, 0.08333333, -0.91666667, 0, 0.66666667, 0.05555556,
-0.61111111, 0, 0.44444444, 0.03703704, -0.40740741
],
8,
err_msg='shat1 should almost equal simple time series, 4th run')
assert_almost_equal(
sighat1, [
0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.70710678
],
8,
err_msg='sighat1 should almost equal simple time series, 4th run')
# smooth three cycles' worth of sinusoid input observations,
# assuming only the seasonal adjustments vary, starting at zero
yobs1 = np.concatenate((s0, s0, s0))
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(yobs1,
m,
alpha=0,
beta=0,
gamma=gamma,
s0=s0 * 0,
l0=0,
b0=0,
sigma0=sigma0,
hstep=hstep)
# check output
assert_almost_equal(
yhat1, [
0, 0, 0, 0, 0, 0.33333333, 0, -0.33333333, 0, 0.55555556, 0,
-0.55555556
],
8,
err_msg='yhat1 should almost equal simple time series, 5th run')
assert_almost_equal(
shat1, [
0, 0, -0.08333333, -0.08333333, 0, 0.33333333, -0.05555556,
-0.38888889, 0, 0.55555555, -0.03703704, -0.59259259
],
8,
err_msg='shat1 should almost equal simple time series, 5th run')
assert_almost_equal(
sighat1, [
0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.70710678
],
8,
err_msg='sighat1 should almost equal simple time series, 5th run')
def test_sqdistalgorithm_additive2():
"""SqDistAlgorithm_test.test_sqdistalgorithm_additive2()
Uses synthetic data time series over 300 days to test additive method
outputs.
"""
# set up smoothing parameters
m = 100 # length of "day"
alpha = 1.0 / 100.0 / 3.0 # average age of level is 3 "days"
beta = 0 # slope doesn't change
gamma = 1.0 / 100.0 * 100.0 / 3.0 # average age of "seasonal" correction
phi = 1 # don't dampen the slope
# initialize states for smoother
l0 = None # this uses the default initial level
b0 = 0 # this is NOT the default initial slope
s0 = None # this uses default initial "seasonal" correction
sigma0 = [0] # this is NOT the default initial standard deviation
# create first 50 "days" at 100 samples per synthetic "day", 0-50
t000to050 = np.arange(5001)
syn000to050 = 10.0 * np.sin(t000to050 * (2 * np.pi) / 100.0)
# these are the old "defaults" computed when l0, b0, and s0 were None
l0 = np.nanmean(syn000to050[0:1440])
b0 = (np.nanmean(syn000to050[m:2 * m]) - np.nanmean(syn000to050[0:m])) / m
s0 = [syn000to050[i] - l0 for i in range(m)]
# run additive method on first 50 "days"
(synHat000to050, sHat000to050, sigma000to050, syn050, s050, l050, b050,
sigma050) = sq.additive(syn000to050,
m,
alpha,
beta,
gamma,
phi,
yhat0=None,
s0=s0,
l0=l0,
b0=b0,
sigma0=sigma0)
# The output should track the input exactly on this simple series
assert_equals(synHat000to050.all(), syn000to050.all(),
'Output of additive should match simple sinusoid exactly')
# Check max, min and average
assert_almost_equal(np.amax(synHat000to050), 10, 8,
'Additive output should have a max of 10.0')
assert_almost_equal(np.amin(synHat000to050), -10, 8,
'Additive output should have a min of -10.0')
assert_almost_equal(np.mean(synHat000to050), 0, 8,
'Additive output should have average of 0')
# create 2nd set of 50 "days", 50-100
t050to100 = np.arange(5001, 10001)
syn050to100 = 20 + 10.0 * np.sin(t050to100 * (2 * np.pi) / 100.0)
# run additive method on next 50 "days"
(synHat050to100, sHat050to100, sigma050to100, syn100, s100, l100, b100,
sigma100) = sq.additive(syn050to100,
m,
alpha,
beta,
gamma,
phi,
yhat0=syn050,
s0=s050,
l0=l050,
b0=b050,
sigma0=sigma050)
# Check max, min and average
assert_almost_equal(np.amax(synHat050to100), 30, 6,
'Additive output should have a max of 30.0')
assert_almost_equal(np.amin(synHat050to100), -8.81753802428088, 8,
'Additive output should have a min of -8.81753...')
assert_almost_equal(np.mean(synHat050to100), 19.17899833054862, 8,
'Additive output should have average of 19.17899...')
# the initial part of the computed series is catching up to the synthetic
assert_array_less(synHat050to100[:555], syn050to100[:555],
'Output of additive should begin below synthetic data')
# short section where the series' swap places
assert_array_less(syn050to100[555:576], synHat050to100[555:576])
# they swap back
assert_array_less(synHat050to100[576:655], syn050to100[576:655])
# swap again
assert_array_less(syn050to100[655:689], synHat050to100[655:689])
# after the initial lag and swaps, the series' get closer and closer
assert_allclose(
syn050to100[475:],
synHat050to100[475:],
rtol=1e-1,
err_msg='Additive output should trend toward synthetic data, 1e-1')
assert_allclose(
syn050to100[955:],
synHat050to100[955:],
rtol=1e-2,
err_msg='Additive output should trend toward synthetic data, 1e-2')
assert_allclose(
syn050to100[1500:],
synHat050to100[1500:],
rtol=1e-3,
err_msg='Additive output should trend toward synthetic data, 1e-3')
assert_allclose(
syn050to100[2100:],
synHat050to100[2100:],
rtol=1e-4,
err_msg='Additive output should trend toward synthetic data, 1e-4')
assert_allclose(
syn050to100[2700:],
synHat050to100[2700:],
rtol=1e-5,
err_msg='Additive output should trend toward synthetic data, 1e-5')
assert_allclose(
syn050to100[3300:],
synHat050to100[3300:],
rtol=1e-6,
err_msg='Additive output should track synthetic data, 1e-6: 50-100')
# create 3rd set of 50 "days", 100-150
t100to150 = np.arange(10001, 15001)
syn100to150 = 20 + 10.0 * np.sin(t100to150 * (2 * np.pi) / 100.) + \
20 * np.sin(t100to150 * (2 * np.pi) / 5000.0)
# run the additive method on the 3rd set of 50 "days"
(synHat100to150, sHat100to150, sigma100to150, syn150, s150, l150, b150,
sigma150) = sq.additive(syn100to150,
m,
alpha,
beta,
gamma,
phi,
yhat0=syn100,
l0=l100,
b0=b100,
s0=s100,
sigma0=sigma100)
# Check max, min and average
assert_almost_equal(np.amax(synHat100to150), 49.758884882080558, 8,
'Additive output should have a max of 49.75888...')
assert_almost_equal(np.amin(synHat100to150), -9.7579516919427647, 8,
'Additive output should have a min of -9.7579...')
assert_almost_equal(np.mean(synHat100to150), 20.059589538984323, 8,
'Additive output should have average of 20.0595...')
# A couple of sections run pretty close together here
assert_allclose(
syn100to150[800:1900],
synHat100to150[800:1900],
rtol=1e-1,
err_msg='Additive output should track synthetic data: day 100-150')
# create 4th set of 50 "days", 150-200
t150to200 = np.arange(15001, 20001)
syn150to200 = 20 + (10.0 * np.sin(t150to200 * (2 * np.pi) / 100.0)) * \
(1 * np.cos(t150to200 * (2 * np.pi) / 5000.0))
# run the additive method on the 4th set of 50 "days"
(synHat150to200, sHat150to200, sigma150to200, syn200, s200, l200, b200,
sigma200) = sq.additive(syn150to200,
m,
alpha,
beta,
gamma,
phi,
yhat0=syn150,
l0=l150,
b0=b150,
s0=s150,
sigma0=sigma150)
# Check max, min and average
assert_almost_equal(np.amax(synHat150to200), 29.573654766341747, 8,
'Additive output should have a max of 29.5736...')
assert_almost_equal(np.amin(synHat150to200), 7.9430807703401669, 8,
'Additive output should have a min of 7.943...')
assert_almost_equal(np.mean(synHat150to200), 19.911560325896119, 8,
'Additive output should have average of 19.911...')
# create 5th set of 50 "days", 200-250
t200to250 = np.arange(20001, 25001)
syn200to250 = 20 + (
(10.0 * np.sin(t200to250 * (2 * np.pi) / 100.0)) *
(1 * np.cos(t200to250 *
(2 * np.pi) / 5000.0)) + 20 * np.sin(t200to250 *
(2 * np.pi) / 5000.0))
# run the additive method on the 5th set of 50 "days"
(synHat200to250, sHat200to250, sigma200to250, syn250, s250, l250, b250,
sigma250) = sq.additive(syn200to250,
m,
alpha,
beta,
gamma,
phi,
yhat0=syn200,
l0=l200,
b0=b200,
s0=s200,
sigma0=sigma200)
# Check max, min and average
assert_almost_equal(np.amax(synHat200to250), 43.417782188651529, 8,
'Additive output should have a max of 43.417...')
assert_almost_equal(np.amin(synHat200to250), -3.4170071669726791, 8,
'Additive output should have a min of -3.417...')
assert_almost_equal(np.mean(synHat200to250), 20.09191068952186, 8,
'Additive output should have average of 20.0919...')
# create 5th set of 50 "days", 250-300
t250to300 = np.arange(25001, 30001)
np.random.seed(123456789)
syn250to300 = 20 + ((10.0 * np.sin(t250to300 * (2 * np.pi) / 100.0)) *
(1 * np.cos(t250to300 * (2 * np.pi) / 5000.0)) +
20 * np.sin(t250to300 * (2 * np.pi) / 5000.0)) + \
5 * np.random.randn(5000)
# run the additive method on the 5th set of 50 "days"
(synHat250to300, sHat250to300, sigma250to300, syn300, s300, l300, b300,
sigma300) = sq.additive(syn250to300,
m,
alpha,
beta,
gamma,
phi,
yhat0=syn250,
l0=l250,
b0=b250,
s0=s250,
sigma0=sigma250)
# Check max, min and average
assert_almost_equal(np.amax(synHat250to300), 49.3099797861343534, 8,
'Additive output should have a max of 49.309...')
assert_almost_equal(np.amin(synHat250to300), -8.7531069723345301, 8,
'Additive output should have a min of -8.783...')
assert_almost_equal(np.mean(synHat250to300), 20.006498585824623, 8,
'Additive output should have average of 20.006...')
def test_sqdistalgorithm_additive1():
"""SqDistAlgorithm_test.test_sqdistalgorithm_additive1()
Uses a simple 12 point data series to compare additive inputs with
corresponding outputs.
"""
# configure to test zero-step predictions of 4 "season" cycles
m = 4
t = np.linspace(0, 2 * np.pi, m + 1)[:-1]
hstep = 0
# initial slope is 0; average age is infinite
b0 = 0
beta = 1 / np.inf
# initial trendline is 0; average age is 12 steps
l0 = 0
alpha = 1 / 12.0
# initial seasonal correction is sinusoid; average age is 12 steps
s0 = np.sin(t)[0:4]
gamma = 1 / 12.0 * m
# standard deviation of unit-amplitude sinusoid
sigma0 = [np.sqrt(0.5)]
# predict three cycles ahead given l0 and s0, no inputs,
# and assume PI only grows with trendline adjustments
yobs1 = np.zeros(12) * np.nan
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(
yobs1, m, alpha=alpha, beta=beta, gamma=0,
s0=s0, l0=l0, b0=b0, sigma0=sigma0, hstep=hstep)
assert_almost_equal(yhat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
err_msg='yhat1 should almost equal simple time series')
assert_almost_equal(shat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
err_msg='shat1 should almost equal simple time series')
assert_almost_equal(sighat1, [0.70710678, 0.70955777, 0.71200031,
0.71443451, 0.71686044, 0.71927819, 0.72168784, 0.72408947, 0.72648316,
0.72886899, 0.73124703, 0.73361737],
err_msg='sighat1 should almost equal simple time series')
# predict three cycles ahead given l0 and s0, no inputs,
# and assume PI only grows with seasonal adjustments
yobs1 = np.zeros(12) * np.nan
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(
yobs1, m, alpha=0, beta=0, gamma=gamma,
s0=s0, l0=0, b0=0, sigma0=sigma0, hstep=hstep)
assert_almost_equal(yhat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
err_msg='yhat1 should almost equal simple time series, 2nd run')
assert_almost_equal(shat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1],
err_msg='shat1 should almost equal simple time series, 2nd run')
assert_almost_equal(sighat1, [0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.74535599, 0.74535599, 0.74535599, 0.74535599, 0.78173596,
0.78173596, 0.78173596, 0.78173596],
err_msg='sighat1 should almost equal simple time series, 2nd run')
# smooth three cycles' worth of zero-value input observations,
# assuming only the trendline varies
yobs1 = np.zeros(12)
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(
yobs1, m, alpha=alpha, beta=0, gamma=0,
s0=s0, l0=0, b0=0, sigma0=sigma0, hstep=hstep)
# check output
assert_almost_equal(yhat1, [0, 1, -0.08333333, -1.07638889, 0.01331019,
1.01220100, -0.07214908, -1.06613666, 0.02270806, 1.02081573,
-0.06425225, -1.0588979], 8,
err_msg='yhat1 should almost equal simple time series, 3rd run')
assert_almost_equal(shat1, [0, 1, 0, -1, 0, 1, 0, -1, 0, 1, 0, -1], 8,
err_msg='shat1 should almost equal simple time series, 3rd run')
assert_almost_equal(sighat1, [0.64818122, 0.67749945, 0.62798561,
0.66535255, 0.61101568, 0.64444779, 0.59675623, 0.63587127, 0.58477433,
0.62111112, 0.57470621, 0.61505552], 8,
err_msg='sighat1 should almost equal simple time series, 3rd run')
# smooth three cycles' worth of zero-value input observations,
# assuming only the seasonal adjustments vary
yobs1 = np.zeros(12)
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(
yobs1, m, alpha=0, beta=0, gamma=gamma,
s0=s0, l0=0, b0=0, sigma0=sigma0, hstep=hstep)
# check output
assert_almost_equal(yhat1, [0, 1, 0, -1, 0, 0.66666667, 0, -0.66666667,
0, 0.44444444, 0, -0.44444444], 8,
err_msg='yhat1 should almost equal simple time series, 4th run')
assert_almost_equal(shat1, [0, 1, 0.08333333, -0.91666667, 0, 0.66666667,
0.05555556, -0.61111111, 0, 0.44444444, 0.03703704, -0.40740741], 8,
err_msg='shat1 should almost equal simple time series, 4th run')
assert_almost_equal(sighat1, [0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.70710678, 0.70710678], 8,
err_msg='sighat1 should almost equal simple time series, 4th run')
# smooth three cycles' worth of sinusoid input observations,
# assuming only the seasonal adjustments vary, starting at zero
yobs1 = np.concatenate((s0, s0, s0))
yhat1, shat1, sighat1, _, _, _, _, _ = sq.additive(
yobs1, m, alpha=0, beta=0, gamma=gamma,
s0=s0 * 0, l0=0, b0=0, sigma0=sigma0, hstep=hstep)
# check output
assert_almost_equal(yhat1, [0, 0, 0, 0, 0, 0.33333333, 0, -0.33333333,
0, 0.55555556, 0, -0.55555556], 8,
err_msg='yhat1 should almost equal simple time series, 5th run')
assert_almost_equal(shat1, [0, 0, -0.08333333, -0.08333333, 0, 0.33333333,
-0.05555556, -0.38888889, 0, 0.55555555, -0.03703704, -0.59259259], 8,
err_msg='shat1 should almost equal simple time series, 5th run')
assert_almost_equal(sighat1, [0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.70710678, 0.70710678,
0.70710678, 0.70710678, 0.70710678], 8,
err_msg='sighat1 should almost equal simple time series, 5th run')
out_stream = Stream()
# loop over channels to retrieve input data for SqDistAlgorithm
for ch in channels:
if state_dir and state_file:
sf = state_dir + state_file + '_' + ob + ch
else:
sf = None
# create SqDistAlgorithm object
svsqdist[ch] = SqDistAlgorithm(
# non-default configuration parameters
alpha=alpha,
beta=beta,
gamma=gamma,
phi=phi,
m=m,
zthresh=zthresh,
statefile=sf)
# load statefile if it exists
svsqdist[ch].load_state()
# set out_start to the next_starttime if it exists
out_start = end_now
out_end = end_now
if (svsqdist[ch].next_starttime is not None and starttime is None
and endtime is None):
out_start = svsqdist[ch].next_starttime
else:
def test_sqdistalgorithm_additive2():
"""SqDistAlgorithm_test.test_sqdistalgorithm_additive2()
Uses synthetic data time series over 300 days to test additive method
outputs.
"""
# set up smoothing parameters
m = 100 # length of "day"
alpha = 1.0 / 100.0 / 3.0 # average age of level is 3 "days"
beta = 0 # slope doesn't change
gamma = 1.0 / 100.0 * 100.0 / 3.0 # average age of "seasonal" correction
phi = 1 # don't dampen the slope
# initialize states for smoother
l0 = None # this uses the default initial level
b0 = 0 # this is NOT the default initial slope
s0 = None # this uses default initial "seasonal" correction
sigma0 = [0] # this is NOT the default initial standard deviation
# create first 50 "days" at 100 samples per synthetic "day", 0-50
t000to050 = np.arange(5001)
syn000to050 = 10.0 * np.sin(t000to050 * (2 * np.pi) / 100.0)
# these are the old "defaults" computed when l0, b0, and s0 were None
l0 = np.nanmean(syn000to050[0:1440])
b0 = (np.nanmean(syn000to050[m:2 * m]) - np.nanmean(syn000to050[0:m])) / m
s0 = [syn000to050[i] - l0 for i in range(m)]
# run additive method on first 50 "days"
(synHat000to050, sHat000to050, sigma000to050,
syn050, s050, l050, b050, sigma050) = sq.additive(
syn000to050, m, alpha, beta, gamma, phi,
yhat0=None, s0=s0, l0=l0, b0=b0, sigma0=sigma0)
# The output should track the input exactly on this simple series
assert_equals(synHat000to050.all(), syn000to050.all(),
'Output of additive should match simple sinusoid exactly')
# Check max, min and average
assert_almost_equal(np.amax(synHat000to050), 10, 8,
'Additive output should have a max of 10.0')
assert_almost_equal(np.amin(synHat000to050), -10, 8,
'Additive output should have a min of -10.0')
assert_almost_equal(np.mean(synHat000to050), 0, 8,
'Additive output should have average of 0')
# create 2nd set of 50 "days", 50-100
t050to100 = np.arange(5001, 10001)
syn050to100 = 20 + 10.0 * np.sin(t050to100 * (2 * np.pi) / 100.0)
# run additive method on next 50 "days"
(synHat050to100, sHat050to100, sigma050to100,
syn100, s100, l100, b100, sigma100) = sq.additive(
syn050to100, m, alpha, beta, gamma, phi,
yhat0=syn050, s0=s050, l0=l050, b0=b050, sigma0=sigma050)
# Check max, min and average
assert_almost_equal(np.amax(synHat050to100), 30, 6,
'Additive output should have a max of 30.0')
assert_almost_equal(np.amin(synHat050to100), -8.81753802428088, 8,
'Additive output should have a min of -8.81753...')
assert_almost_equal(np.mean(synHat050to100), 19.17899833054862, 8,
'Additive output should have average of 19.17899...')
# the initial part of the computed series is catching up to the synthetic
assert_array_less(synHat050to100[:555], syn050to100[:555],
'Output of additive should begin below synthetic data')
# short section where the series' swap places
assert_array_less(syn050to100[555:576], synHat050to100[555:576])
# they swap back
assert_array_less(synHat050to100[576:655], syn050to100[576:655])
# swap again
assert_array_less(syn050to100[655:689], synHat050to100[655:689])
# after the initial lag and swaps, the series' get closer and closer
assert_allclose(syn050to100[475:], synHat050to100[475:], rtol=1e-1,
err_msg='Additive output should trend toward synthetic data, 1e-1')
assert_allclose(syn050to100[955:], synHat050to100[955:], rtol=1e-2,
err_msg='Additive output should trend toward synthetic data, 1e-2')
assert_allclose(syn050to100[1500:], synHat050to100[1500:], rtol=1e-3,
err_msg='Additive output should trend toward synthetic data, 1e-3')
assert_allclose(syn050to100[2100:], synHat050to100[2100:], rtol=1e-4,
err_msg='Additive output should trend toward synthetic data, 1e-4')
assert_allclose(syn050to100[2700:], synHat050to100[2700:], rtol=1e-5,
err_msg='Additive output should trend toward synthetic data, 1e-5')
assert_allclose(syn050to100[3300:], synHat050to100[3300:], rtol=1e-6,
err_msg='Additive output should track synthetic data, 1e-6: 50-100')
# create 3rd set of 50 "days", 100-150
t100to150 = np.arange(10001, 15001)
syn100to150 = 20 + 10.0 * np.sin(t100to150 * (2 * np.pi) / 100.) + \
20 * np.sin(t100to150 * (2 * np.pi) / 5000.0)
# run the additive method on the 3rd set of 50 "days"
(synHat100to150, sHat100to150, sigma100to150,
syn150, s150, l150, b150, sigma150) = sq.additive(
syn100to150, m, alpha, beta, gamma, phi,
yhat0=syn100, l0=l100, b0=b100, s0=s100, sigma0=sigma100)
# Check max, min and average
assert_almost_equal(np.amax(synHat100to150), 49.758884882080558, 8,
'Additive output should have a max of 49.75888...')
assert_almost_equal(np.amin(synHat100to150), -9.7579516919427647, 8,
'Additive output should have a min of -9.7579...')
assert_almost_equal(np.mean(synHat100to150), 20.059589538984323, 8,
'Additive output should have average of 20.0595...')
# A couple of sections run pretty close together here
assert_allclose(syn100to150[800:1900], synHat100to150[800:1900], rtol=1e-1,
err_msg='Additive output should track synthetic data: day 100-150')
# create 4th set of 50 "days", 150-200
t150to200 = np.arange(15001, 20001)
syn150to200 = 20 + (10.0 * np.sin(t150to200 * (2 * np.pi) / 100.0)) * \
(1 * np.cos(t150to200 * (2 * np.pi) / 5000.0))
# run the additive method on the 4th set of 50 "days"
(synHat150to200, sHat150to200, sigma150to200,
syn200, s200, l200, b200, sigma200) = sq.additive(
syn150to200, m, alpha, beta, gamma, phi,
yhat0=syn150, l0=l150, b0=b150, s0=s150, sigma0=sigma150)
# Check max, min and average
assert_almost_equal(np.amax(synHat150to200), 29.573654766341747, 8,
'Additive output should have a max of 29.5736...')
assert_almost_equal(np.amin(synHat150to200), 7.9430807703401669, 8,
'Additive output should have a min of 7.943...')
assert_almost_equal(np.mean(synHat150to200), 19.911560325896119, 8,
'Additive output should have average of 19.911...')
# create 5th set of 50 "days", 200-250
t200to250 = np.arange(20001, 25001)
syn200to250 = 20 + ((10.0 * np.sin(t200to250 * (2 * np.pi) / 100.0)) *
(1 * np.cos(t200to250 * (2 * np.pi) / 5000.0)) +
20 * np.sin(t200to250 * (2 * np.pi) / 5000.0))
# run the additive method on the 5th set of 50 "days"
(synHat200to250, sHat200to250, sigma200to250,
syn250, s250, l250, b250, sigma250) = sq.additive(
syn200to250, m, alpha, beta, gamma, phi,
yhat0=syn200, l0=l200, b0=b200, s0=s200, sigma0=sigma200)
# Check max, min and average
assert_almost_equal(np.amax(synHat200to250), 43.417782188651529, 8,
'Additive output should have a max of 43.417...')
assert_almost_equal(np.amin(synHat200to250), -3.4170071669726791, 8,
'Additive output should have a min of -3.417...')
assert_almost_equal(np.mean(synHat200to250), 20.09191068952186, 8,
'Additive output should have average of 20.0919...')
# create 5th set of 50 "days", 250-300
t250to300 = np.arange(25001, 30001)
np.random.seed(123456789)
syn250to300 = 20 + ((10.0 * np.sin(t250to300 * (2 * np.pi) / 100.0)) *
(1 * np.cos(t250to300 * (2 * np.pi) / 5000.0)) +
20 * np.sin(t250to300 * (2 * np.pi) / 5000.0)) + \
5 * np.random.randn(5000)
# run the additive method on the 5th set of 50 "days"
(synHat250to300, sHat250to300, sigma250to300,
syn300, s300, l300, b300, sigma300) = sq.additive(
syn250to300, m, alpha, beta, gamma, phi,
yhat0=syn250, l0=l250, b0=b250, s0=s250, sigma0=sigma250)
# Check max, min and average
assert_almost_equal(np.amax(synHat250to300), 49.3099797861343534, 8,
'Additive output should have a max of 49.309...')
assert_almost_equal(np.amin(synHat250to300), -8.7531069723345301, 8,
'Additive output should have a min of -8.783...')
assert_almost_equal(np.mean(synHat250to300), 20.006498585824623, 8,
'Additive output should have average of 20.006...')