#--------------------------------
# Compute EWS without bootstrapping
#-------------------------------------
# Time-series data as a pandas Series
series = df_traj['x']
# Put into ews_compute
ews_dic = ews_compute(series,
smooth = 'Lowess',
span = span,
roll_window = rw,
upto = tcrit,
ews = ews,
lag_times = lags,
sweep = sweep,
ham_length=ham_length,
ham_offset=ham_offset)
# DataFrame of EWS
df_ews = ews_dic['EWS metrics']
# DataFrame of Power spectra
df_pspec = ews_dic['Power spectrum']
# Plot trajectory and smoothing
df_ews[['State variable','Smoothing']].plot()
appended_ews = []
# Print update
print('\nBegin EWS computation\n')
# Loop through noise type
for noise_type in ['add', 'multi', 'k']:
# Loop through h value
for h in hVals:
ews_dic = ews_compute(df_traj_filt.loc[noise_type].loc[h]['State'],
roll_window = rw,
band_width = bw,
lag_times = lags,
ews = ews,
ham_length = ham_length,
ham_offset = ham_offset,
pspec_roll_offset = pspec_roll_offset
)
# The DataFrame of EWS
df_ews_temp = ews_dic['EWS metrics']
# Include a column in the DataFrames for noise type and h value
df_ews_temp['h'] = h
df_ews_temp['Noise'] = noise_type
# Add DataFrames to list
appended_ews.append(df_ews_temp)
# set up a list to store output dataframes from ews_compute- we will concatenate them at the end
appended_ews = []
appended_ktau = []
# loop through realisation number
print('\nBegin EWS computation\n')
for i in range(numSims):
# loop through variable
for var in ['x']:
ews_dic = ews_compute(df_traj_filt.loc[i + 1][var],
roll_window=rw,
smooth='Lowess',
span=span,
lag_times=lags,
ews=ews,
ham_length=ham_length,
ham_offset=ham_offset,
pspec_roll_offset=pspec_roll_offset,
upto=tcrit)
# The DataFrame of EWS
df_ews_temp = ews_dic['EWS metrics']
# The DataFrame of kendall tau values
df_ktau_temp = ews_dic['Kendall tau']
# Include a column in the DataFrames for realisation number and variable
df_ews_temp['Realisation number'] = i + 1
df_ews_temp['Variable'] = var
# set up a list to store output dataframes from ews_compute- we will concatenate them at the end
appended_ews = []
appended_pspec = []
# loop through realisation number
print('\nBegin EWS computation\n')
for i in range(numSims):
# loop through variable
for var in ['x', 'y']:
ews_dic = ews_compute(df_traj_filt.loc[i + 1][var],
roll_window=rw,
band_width=bw,
lag_times=lags,
ews=ews,
ham_length=ham_length,
ham_offset=ham_offset,
pspec_roll_offset=pspec_roll_offset,
upto=tbif)
# The DataFrame of EWS
df_ews_temp = ews_dic['EWS metrics']
# The DataFrame of power spectra
df_pspec_temp = ews_dic['Power spectrum']
# Include a column in the DataFrames for realisation number and variable
df_ews_temp['Realisation number'] = i + 1
df_ews_temp['Variable'] = var
df_pspec_temp['Realisation number'] = i + 1
#--------------------------------
# Compute EWS (moments) without bootstrapping
#-------------------------------------
# Filter time-series to have time-spacing dt2
df_traj_filt = df_traj.loc[::int(dt2/dt)]
# Time-series data as a pandas Series
series = df_traj_filt['x']
# Put into ews_compute
ews_dic = ews_compute(series,
smooth = 'Lowess',
span = span,
roll_window = rw,
upto = tcrit,
ews = ews)
# DataFrame of EWS
df_ews = ews_dic['EWS metrics']
# Plot trajectory and smoothing
df_ews[['State variable','Smoothing']].plot()
# Plot variance
df_ews[['Variance']].plot()
# Trajectory data stored in a DataFrame indexed by time
data = {'Time': t, 'x': x}
df_traj = pd.DataFrame(data).set_index('Time')
#--------------------------------
# Compute EWS (moments) without bootstrapping
#-------------------------------------
# Time-series data as a pandas Series
series = df_traj['x']
# Put into ews_compute
ews_dic = ews_compute(series,
smooth='Lowess',
span=span,
roll_window=0.5,
upto=tcrit,
ews=ews)
# DataFrame of EWS
df_ews = ews_dic['EWS metrics']
# Plot trajectory and smoothing
df_ews[['State variable', 'Smoothing']].plot()
# Plot variance
df_ews[['Variance']].plot()
#-------------------------------------
# Compute EWS using bootstrapping
#–----------------------------------
dt2 = 1
df_sims_filt = df_sims[np.remainder(df_sims.index,dt2) == 0]
# set up a list to store output dataframes from ews_compute- we will concatenate them at the end
appended_ews = []
appended_ktau = []
# Print update
print('\n Begin EWS computation \n')
# loop through each trajectory as an input to ews_compute
for i in range(numSims):
dict_ews = ews_compute(df_sims_filt['Sim '+str(i+1)],
roll_window=0.5,
band_width=0.05,
lag_times=[1],
ews=['var','ac','sd','cv','skew','kurt','smax'],
ham_length=40,
upto=tbif*0.9,
pspec_roll_offset = 20)
# EWS dataframe
df_ews_temp = dict_ews['EWS metrics']
# Include a column in the dataframe for realisation number
df_ews_temp['Realisation number'] = i+1
# # Power spectra dataframe
# df_pspec_temp = dict_ews['Power spectrum']
# df_pspec_temp['Realisation number'] = i+1
# Kendall tau values
df_ktau_temp = dict_ews['Kendall tau']
sampleVals = np.array(df_samples.index.levels[1])
# Loop through realtimes
for t in tVals:
# Loop through sample values
for sample in sampleVals:
# Compute EWS for near-stationary sample series
series_temp = df_samples.loc[t].loc[sample]['x']
ews_dic = ews_compute(series_temp,
roll_window=1,
band_width=1,
lag_times=lags,
ews=ews,
ham_length=ham_length,
ham_offset=ham_offset,
pspec_roll_offset=pspec_roll_offset,
upto=1,
sweep=False)
# The DataFrame of EWS
df_ews_temp = ews_dic['EWS metrics']
# Include columns for sample value and realtime
df_ews_temp['Sample'] = sample
df_ews_temp['Time'] = t
# Drop NaN values
df_ews_temp = df_ews_temp.dropna()
