# scales
scales = wa.scales
# associated time vector
t = wa.time / 24.
# reconstruction of the original data
rx = wa.reconstruction()
# determine acor factor for red noise
acorr = acf(x)
lagf = (acorr[1] + np.sqrt(acorr[2])) / 2
print 'acorr lagf is %s' % lagf
# determine significance levels
(signif, fft_theory) = wave_sig.wave_signif(x, dt, scales, lag1=lagf)
sig95 = np.ones_like(power) * np.array([signif] * len(t)).transpose()
sig95 = power / sig95 # where ratio > 1, power is significant
# Global wavelet spectrum & significance levels:
global_int = variance * (np.sum(
power, axis=0)) / x.shape[0] # time-average over all times
gs = ((np.sum(power, axis=1)) / x.shape[0]) / variance #assume var=1
gswa = wa.global_wavelet_spectrum
# Global wavelet significance
(signif_g, fft_theory_g) = wave_sig.global_wave_signif(x,
dt,
scales,
lag1=lagf,
sigtest=1,
dof=len(x))
# determine acor factor for red noise stream 1
# uses entire dataperiod for corr
#TODO: current implementation uses any preprocessed changes in ingest module (detrended, standardized, etc)
acorr_1 = acf(data_1['anom'])
lagf_1 = (acorr_1[1]+np.sqrt(acorr_1[2]))/2
print 'acorr lagf for datastream 1 is %s' % lagf_1
scalemin = 1
scalemax = 32
scale_ind = ((wa1c.scales >= scalemin) & (wa1c.scales <= scalemax))
# determine significance levels for stream 1
(signif_1, fft_theory_1) = wave_sig.wave_signif(data_1c['anom'],wa1c.dt,wa1c.scales[scale_ind],lag1=lagf_1)
sig95_1 = np.ones_like(wa1c.wavelet_power[scale_ind,:]) * np.array([signif_1] * len(wa1c.time)).transpose()
sig95_1 = wa1c.wavelet_power[scale_ind,:] / sig95_1 # where ratio > 1, power is significant
# Global wavelet spectrum & significance levels stream 1:
global_int_1 = data_1['variance']*(np.sum(wa1c.wavelet_power, axis=0) ) / data_1c['anom'].shape[0] # time-average over all times
gs_1 = ((np.sum(wa1c.wavelet_power, axis=1) ) / data_1c['anom'].shape[0]) / data_1['variance'] #assume var=1
gswa_1 = wa1c.global_wavelet_spectrum
# Global wavelet significance
(signif_g_1, fft_theory_g_1) = wave_sig.global_wave_signif(data_1c['anom'],wa1c.dt,wa1c.scales,lag1=lagf_1,sigtest=1, dof=len(data_1c['anom']))
"""----------------------------- plot setup ------------------------------------------"""
# scales
scales = wa.scales
# associated time vector
t = wa.time / 24.
# reconstruction of the original data
rx = wa.reconstruction()
# determine acor factor for red noise
acorr = acf(x)
lagf = (acorr[1]+np.sqrt(acorr[2]))/2
print 'acorr lagf is %s' % lagf
# determine significance levels
(signif, fft_theory) = wave_sig.wave_signif(x,dt,scales,lag1=lagf)
sig95 = np.ones_like(power) * np.array([signif] * len(t)).transpose()
sig95 = power / sig95 # where ratio > 1, power is significant
# Global wavelet spectrum & significance levels:
global_int = variance*(np.sum(power, axis=0) ) / x.shape[0] # time-average over all times
gs = ((np.sum(power, axis=1) ) / x.shape[0]) / variance #assume var=1
gswa = wa.global_wavelet_spectrum
# Global wavelet significance
(signif_g, fft_theory_g) = wave_sig.global_wave_signif(x,dt,scales,lag1=lagf,sigtest=1, dof=len(x))
"""----------------------------- plot setup ------------------------------------------"""
T, S = np.meshgrid(t, scales)
# determine acor factor for red noise stream 1
# uses entire dataperiod for corr
#TODO: current implementation uses any preprocessed changes in ingest module (detrended, standardized, etc)
acorr_1 = acf(data_1['anom'])
lagf_1 = (acorr_1[1]+np.sqrt(acorr_1[2]))/2
print 'acorr lagf for datastream 1 is %s' % lagf_1
# determine acor factor for red noise stream 2
# uses entire dataperiod for corr
acorr_2 = acf(data_2['anom'])
lagf_2 = (acorr_2[1]+np.sqrt(acorr_2[2]))/2
print 'acorr lagf for datastream 1 is %s' % lagf_2
# determine significance levels for stream 1
(signif_1, fft_theory_1) = wave_sig.wave_signif(data_1c['anom'],wa1c.dt,wa1c.scales,lag1=lagf_1)
sig95_1 = np.ones_like(wa1c.wavelet_power) * np.array([signif_1] * len(wa1c.time)).transpose()
sig95_1 = wa1c.wavelet_power / sig95_1 # where ratio > 1, power is significant
# determine significance levels for stream 1
(signif_2, fft_theory_2) = wave_sig.wave_signif(data_2c['anom'],wa2c.dt,wa2c.scales,lag1=lagf_2)
sig95_2 = np.ones_like(wa2c.wavelet_power) * np.array([signif_2] * len(wa2c.time)).transpose()
sig95_2 = wa2c.wavelet_power / sig95_2 # where ratio > 1, power is significant
# Global wavelet spectrum & significance levels stream 1:
global_int_1 = data_1['variance']*(np.sum(wa1c.wavelet_power, axis=0) ) / data_1c['anom'].shape[0] # time-average over all times
gs_1 = ((np.sum(wa1c.wavelet_power, axis=1) ) / data_1c['anom'].shape[0]) / data_1['variance'] #assume var=1
gswa_1 = wa1c.global_wavelet_spectrum
# Global wavelet significance
(signif_g_1, fft_theory_g_1) = wave_sig.global_wave_signif(data_1c['anom'],wa1c.dt,wa1c.scales,lag1=lagf_1,sigtest=1, dof=len(data_1c['anom']))
# determine acor factor for red noise stream 1
# uses entire dataperiod for corr
#TODO: current implementation uses any preprocessed changes in ingest module (detrended, standardized, etc)
acorr_1 = acf(data_1['anom'])
lagf_1 = (acorr_1[1] + np.sqrt(acorr_1[2])) / 2
print 'acorr lagf for datastream 1 is %s' % lagf_1
scalemin = 1
scalemax = 32
scale_ind = ((wa1c.scales >= scalemin) & (wa1c.scales <= scalemax))
# determine significance levels for stream 1
(signif_1, fft_theory_1) = wave_sig.wave_signif(data_1c['anom'],
wa1c.dt,
wa1c.scales[scale_ind],
lag1=lagf_1)
sig95_1 = np.ones_like(wa1c.wavelet_power[scale_ind, :]) * np.array(
[signif_1] * len(wa1c.time)).transpose()
sig95_1 = wa1c.wavelet_power[
scale_ind, :] / sig95_1 # where ratio > 1, power is significant
# Global wavelet spectrum & significance levels stream 1:
global_int_1 = data_1['variance'] * (np.sum(
wa1c.wavelet_power,
axis=0)) / data_1c['anom'].shape[0] # time-average over all times
gs_1 = ((np.sum(wa1c.wavelet_power, axis=1)) /
data_1c['anom'].shape[0]) / data_1['variance'] #assume var=1
gswa_1 = wa1c.global_wavelet_spectrum
# Global wavelet significance
(signif_g_1, fft_theory_g_1) = wave_sig.global_wave_signif(
# scales
scales = wa.scales
# associated time vector
t = wa.time
# reconstruction of the original data
rx = wa.reconstruction()
# determine acor factor for red noise
acorr = acf(data_1['anom'])
lagf = (acorr[1]+np.sqrt(acorr[2]))/2
print 'acorr lagf is %s' % lagf
# determine significance levels
(signif, fft_theory) = wave_sig.wave_signif(data_1['anom'],data_1['dt'],scales,lag1=lagf)
sig95 = np.ones_like(power) * np.array([signif] * len(t)).transpose()
sig95 = power / sig95 # where ratio > 1, power is significant
# Global wavelet spectrum & significance levels:
global_int = data_1['variance']*(np.sum(power, axis=0) ) / data_1['anom'].shape[0] # time-average over all times
gs = ((np.sum(power, axis=1) ) / data_1['anom'].shape[0]) / data_1['variance'] #assume var=1
gswa = wa.global_wavelet_spectrum
# Global wavelet significance
(signif_g, fft_theory_g) = wave_sig.global_wave_signif(data_1['anom'],data_1['dt'],scales,lag1=lagf,sigtest=1, dof=len(data_1['anom']))
"""----------------------------- plot setup ------------------------------------------"""
T, S = np.meshgrid(t, scales)