def get_probs_copula_props(probs, probs_rolled, n_ecop_bins):
scorr = np.corrcoef(probs, probs_rolled)[0, 1]
asymm_1, asymm_2 = get_asymms_sample(probs, probs_rolled)
asymm_1 /= get_asymm_1_max(scorr)
asymm_2 /= get_asymm_2_max(scorr)
# plt.scatter(probs, probs_rolled, alpha=0.5)
# plt.grid()
# plt.show()
# plt.close()
ecop_dens_arr = np.full((n_ecop_bins, n_ecop_bins),
np.nan,
dtype=np.float64)
fill_bi_var_cop_dens(probs, probs_rolled, ecop_dens_arr)
non_zero_idxs = ecop_dens_arr > 0
dens = ecop_dens_arr[non_zero_idxs]
etpy_arr = -(dens * np.log(dens))
etpy = etpy_arr.sum()
etpy_min = get_etpy_min(n_ecop_bins)
etpy_max = get_etpy_max(n_ecop_bins)
etpy = (etpy - etpy_min) / (etpy_max - etpy_min)
return asymm_1, asymm_2, etpy, scorr
def main():
main_dir = Path(r'P:\Synchronize\IWS\Testings\fourtrans_practice\phsann')
os.chdir(main_dir)
in_file = Path(r'neckar_norm_cop_infill_discharge_1961_2015_20190118.csv')
lag_steps = np.arange(1, 3, dtype=np.int64)
ecop_bins = 20
fig_size = (15, 10)
plt_alpha = 0.5
in_ser = get_data_df(in_file)
idx_labs = np.unique(in_ser.index.year)
etpy_min = _get_etpy_min(ecop_bins)
etpy_max = _get_etpy_max(ecop_bins)
ecop_dens_arrs = np.full((ecop_bins, ecop_bins), np.nan, dtype=np.float64)
axes = plt.subplots(2, 3, squeeze=False, figsize=fig_size)[1]
cmap = 'jet'
sim_clrs = plt.get_cmap(cmap)(
(idx_labs - idx_labs.min()) / (idx_labs.max() - idx_labs.min()))
sim_clrs = {
idx_lab: sim_clr
for (idx_lab, sim_clr) in zip(idx_labs, sim_clrs)
}
cmap_mappable_beta = plt.cm.ScalarMappable(cmap=cmap)
cmap_mappable_beta.set_array([])
for idx_lab in idx_labs:
data = in_ser.loc[f'{idx_lab}-01-01':f'{idx_lab}-12-31'].values
probs = rankdata(data) / (data.size + 1.0)
scorrs = []
asymms_1 = []
asymms_2 = []
etpys = []
pcorrs = []
for lag_step in lag_steps:
probs_i, rolled_probs_i = roll_real_2arrs(probs, probs, lag_step)
data_i, rolled_data_i = roll_real_2arrs(data, data, lag_step)
# scorr.
scorr = np.corrcoef(probs_i, rolled_probs_i)[0, 1]
scorrs.append(scorr)
# asymms.
asymm_1, asymm_2 = get_asymms_sample(probs_i, rolled_probs_i)
asymm_1 /= _get_asymm_1_max(scorr)
asymm_2 /= _get_asymm_2_max(scorr)
asymms_1.append(asymm_1)
asymms_2.append(asymm_2)
# ecop etpy.
fill_bi_var_cop_dens(probs_i, rolled_probs_i, ecop_dens_arrs)
non_zero_idxs = ecop_dens_arrs > 0
dens = ecop_dens_arrs[non_zero_idxs]
etpy_arr = -(dens * np.log(dens))
etpy = etpy_arr.sum()
etpy = (etpy - etpy_min) / (etpy_max - etpy_min)
etpys.append(etpy)
# pcorr.
pcorr = np.corrcoef(data_i, rolled_data_i)[0, 1]
pcorrs.append(pcorr)
# plot
axes[0, 0].plot(lag_steps,
scorrs,
alpha=plt_alpha,
color=sim_clrs[idx_lab],
label=idx_lab)
axes[1, 0].plot(lag_steps,
asymms_1,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
axes[1, 1].plot(lag_steps,
asymms_2,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
axes[0, 1].plot(lag_steps,
etpys,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
axes[0, 2].plot(lag_steps,
pcorrs,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
axes[1, 2].plot(data, alpha=plt_alpha, color=sim_clrs[idx_lab])
axes[0, 0].grid()
axes[1, 0].grid()
axes[1, 1].grid()
axes[0, 1].grid()
axes[0, 2].grid()
axes[1, 2].grid()
axes[0, 0].legend()
axes[0, 0].set_ylabel('Spearman correlation')
axes[1, 0].set_xlabel('Lag steps')
axes[1, 0].set_ylabel('Asymmetry (Type - 1)')
axes[1, 1].set_xlabel('Lag steps')
axes[1, 1].set_ylabel('Asymmetry (Type - 2)')
axes[0, 1].set_ylabel('Entropy')
axes[0, 2].set_xlabel('Lag steps')
axes[0, 2].set_ylabel('Pearson correlation')
# axes[1, 2].set_xlabel('Nth orders')
# axes[1, 2].set_ylabel('Dist. Sum')
# cbaxes = fig.add_axes([0.2, 0.0, 0.65, 0.05])
# plt.colorbar(
# mappable=cmap_mappable_beta,
# cax=axes[1, 2],
# orientation='horizontal',
# label='Relative Timing',
# alpha=plt_alpha,
# drawedges=False)
plt.show()
return
def main():
main_dir = Path(r'P:\Synchronize\IWS\Testings\fourtrans_practice\phsann')
os.chdir(main_dir)
in_file = Path(r'neckar_norm_cop_infill_discharge_1961_2015_20190118.csv')
lag_steps = np.arange(1, 5, dtype=np.int64)
ecop_bins = 20
fig_size = (15, 10)
plt_alpha = 0.5
in_ser = pd.read_csv(in_file, sep=';', index_col=0)['420']
in_ser.index = pd.to_datetime(in_ser.index, format='%Y-%m-%d')
idx_labs = np.unique(in_ser.index.year)
ecop_dens_arr = np.full((ecop_bins, ecop_bins), np.nan, dtype=np.float64)
lag_axes = {
lag_step: plt.subplots(2, 3, squeeze=False, figsize=fig_size)
for lag_step in lag_steps
}
cmap = 'jet'
sim_clrs = plt.get_cmap(cmap)(
(idx_labs - idx_labs.min()) / (idx_labs.max() - idx_labs.min()))
sim_clrs = {
idx_lab: sim_clr
for (idx_lab, sim_clr) in zip(idx_labs, sim_clrs)
}
cmap_mappable_beta = plt.cm.ScalarMappable(cmap=cmap)
cmap_mappable_beta.set_array([])
for idx_lab in idx_labs:
data = in_ser.loc[f'{idx_lab}-01-01':f'{idx_lab+5}-12-31'].values
probs = rankdata(data) / (data.size + 1.0)
# scorrs = []
# asymms_1 = []
# asymms_2 = []
# etpys = []
# pcorrs = []
# cdf_valss = []
# etpy_cdf_valss = []
for lag_step in lag_steps:
probs_i, rolled_probs_i = roll_real_2arrs(probs, probs, lag_step)
data_i, rolled_data_i = roll_real_2arrs(data, data, lag_step)
# scorr.
scorrs = np.sort(rolled_probs_i * probs_i)
# asymms.
asymms_1 = np.sort((probs_i + rolled_probs_i - 1.0)**3)
asymms_2 = np.sort((probs_i - rolled_probs_i)**3)
# ecop etpy.
fill_bi_var_cop_dens(probs_i, rolled_probs_i, ecop_dens_arr)
non_zero_idxs = ecop_dens_arr > 0
dens = ecop_dens_arr[non_zero_idxs]
etpy_arr = -(dens * np.log(dens))
etpys = np.sort(etpy_arr)
# pcorr.
pcorrs = np.sort((rolled_data_i - data_i))
# cdf vals.
cdf_vals = np.arange(1.0, data_i.size + 1)
cdf_vals /= cdf_vals.size + 1.0
# etpy cdf_vals
etpy_cdf_vals = np.arange(1.0, etpy_arr.size + 1)
etpy_cdf_vals /= etpy_cdf_vals.size + 1.0
# plot
plt.figure(lag_axes[lag_step][0].number)
lag_axes[lag_step][1][0, 0].plot(scorrs,
cdf_vals,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
lag_axes[lag_step][1][1, 0].plot(asymms_1,
cdf_vals,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
lag_axes[lag_step][1][1, 1].plot(asymms_2,
cdf_vals,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
lag_axes[lag_step][1][0, 1].plot(etpys,
etpy_cdf_vals,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
lag_axes[lag_step][1][0, 2].plot(pcorrs,
cdf_vals,
alpha=plt_alpha,
color=sim_clrs[idx_lab])
lag_axes[lag_step][1][0, 0].grid()
lag_axes[lag_step][1][1, 0].grid()
lag_axes[lag_step][1][1, 1].grid()
lag_axes[lag_step][1][0, 1].grid()
lag_axes[lag_step][1][0, 2].grid()
lag_axes[lag_step][1][1, 2].grid()
lag_axes[lag_step][1][0, 0].set_ylabel('Spearman correlation')
lag_axes[lag_step][1][1, 0].set_xlabel('Lag steps')
lag_axes[lag_step][1][1, 0].set_ylabel('Asymmetry (Type - 1)')
lag_axes[lag_step][1][1, 1].set_xlabel('Lag steps')
lag_axes[lag_step][1][1, 1].set_ylabel('Asymmetry (Type - 2)')
lag_axes[lag_step][1][0, 1].set_ylabel('Entropy')
lag_axes[lag_step][1][0, 2].set_xlabel('Lag steps')
lag_axes[lag_step][1][0, 2].set_ylabel('Pearson correlation')
# lag_axes[lag_step][1][1, 2].set_xlabel('Nth orders')
# lag_axes[lag_step][1][1, 2].set_ylabel('Dist. Sum')
plt.colorbar(mappable=cmap_mappable_beta,
cax=lag_axes[lag_step][1][1, 2],
orientation='horizontal',
label='Relative Timing',
alpha=plt_alpha,
drawedges=False)
lag_axes[lag_step][0].suptitle(f'lag_step:{lag_step}')
for lag_step in lag_steps:
plt.figure(lag_axes[lag_step][0].number)
plt.show(block=True)
print(lag_step)
return