def plot_pairwise(self, par_names=None, prior_info_csv_file=None, fig_filename='pairwise_correlation.png',
figure_size=(10, 10)):
""" creates pairwise correlation between parameters specified by ids
:param par_names: (list) names of parameter to display
:param prior_info_csv_file: (string) filename where parameter prior ranges are located
(Note: '!' will be replaced with '\n')
:param fig_filename: (string) filename to save the figure as
:param figure_size: (tuple) figure size
"""
# read information about prior distributions
dict_of_priors = None
if prior_info_csv_file is not None:
dict_of_priors = self.get_dict_of_priors(prior_info_csv_file=prior_info_csv_file)
# if parameter names are not specified, include all parameters
if par_names is None:
par_names = self.get_all_parameter_names()
# find the info of parameters to include in the analysis
info_of_param_info_to_include = []
# for each parameter, read sampled parameter values and create the histogram
for par_name in par_names:
# get parameter values
par_values = self.dictOfParamValues[par_name]
# get info
title, multiplier, x_range, decimal, form = self.get_title_multiplier_x_range_decimal_format(
par_name=par_name, dict_of_priors=dict_of_priors)
# adjust parameter values
par_values = [v*multiplier for v in par_values]
# append the info for this parameter
info_of_param_info_to_include.append(
ParamInfo(name=par_name,
label=title.replace('!', '\n'),
values=par_values, value_range=x_range))
# plot pairwise
# set default properties of the figure
plt.rc('font', size=6) # fontsize of texts
plt.rc('axes', titlesize=6) # fontsize of the figure title
plt.rc('axes', titleweight='semibold') # fontweight of the figure title
# plot each panel
n = len(info_of_param_info_to_include)
if n == 0:
raise ValueError('Values of parameters are not provided. '
'Make sure the calibration algorithm exports the parameter values.')
f, axarr = plt.subplots(nrows=n, ncols=n, figsize=figure_size)
for i in range(n):
for j in range(n):
# get the current axis
ax = axarr[i, j]
if j == 0:
ax.set_ylabel(info_of_param_info_to_include[i].label)
if i == n-1:
ax.set_xlabel(info_of_param_info_to_include[j].label)
if i == j:
# plot histogram
Fig.add_histogram_to_ax(
ax=ax,
data=info_of_param_info_to_include[i].values,
x_range=info_of_param_info_to_include[i].range
)
ax.set_yticklabels([])
ax.set_yticks([])
else:
ax.scatter(info_of_param_info_to_include[j].values,
info_of_param_info_to_include[i].values,
alpha=0.5, s=2)
ax.set_xlim(info_of_param_info_to_include[j].range)
ax.set_ylim(info_of_param_info_to_include[i].range)
# correlation line
b, m = polyfit(info_of_param_info_to_include[j].values,
info_of_param_info_to_include[i].values, 1)
ax.plot(info_of_param_info_to_include[j].values,
b + m * info_of_param_info_to_include[j].values, '-', c='black')
corr, p = pearsonr(info_of_param_info_to_include[j].values,
info_of_param_info_to_include[i].values)
ax.text(0.95, 0.95, '{0:.2f}'.format(corr), transform=ax.transAxes, fontsize=6,
va='top', ha='right')
f.align_ylabels(axarr[:, 0])
f.tight_layout()
output_figure(plt=f, filename=fig_filename, dpi=300)
def plot_pairwise(self,
ids=None,
csv_file_name_prior=None,
fig_filename='pairwise_correlation.png',
figure_size=(10, 10)):
""" creates pairwise corrolation between parameters specified by ids
:param ids: (list) list of parameter ids
:param csv_file_name_prior: (string) filename where parameter prior ranges are located
:param fig_filename: (string) filename to save the figure as
:param figure_size: (tuple) figure size
"""
# read prior distributions
priors = None
if csv_file_name_prior is not None:
priors = IO.read_csv_rows(file_name=csv_file_name_prior,
if_ignore_first_row=True,
delimiter=',',
if_convert_float=True)
# find the names of parameters to include in the analysis
info_of_params_to_include = []
par_id = 0
for key, par_values in self.dictOfParams.items():
# skip these columns
if key in ['Simulation Replication', 'Random Seed']:
continue
# check if the histogram should be created for this parameter
if_show = False
if ids is None:
if_show = True
elif par_id in ids:
if_show = True
# create the histogram
if if_show:
# find prior range
x_range = None
if priors is not None:
try:
x_range = [
float(priors[par_id][Column.LB.value]),
float(priors[par_id][Column.UB.value])
]
except:
print(
'Could not convert string to float to find the prior distribution of parameter:',
par_id)
else:
x_range = None
# find title
if priors[par_id][Column.TITLE.value] in ('', None):
label = priors[par_id][Column.NAME.value]
else:
label = priors[par_id][Column.TITLE.value]
# find multiplier
if priors[par_id][Column.MULTIPLIER.value] in ('', None):
multiplier = 1
else:
multiplier = float(priors[par_id][Column.MULTIPLIER.value])
x_range = [x * multiplier for x in x_range]
par_values = [v * multiplier for v in par_values]
# append the info for this parameter
info_of_params_to_include.append(
ParamInfo(idx=par_id,
name=key,
label=label.replace('!', '\n'),
values=par_values,
range=x_range))
# move to the next parameter
par_id += 1
# plot pairwise
# set default properties of the figure
plt.rc('font', size=6) # fontsize of texts
plt.rc('axes', titlesize=6) # fontsize of the figure title
plt.rc('axes',
titleweight='semibold') # fontweight of the figure title
# plot each panel
n = len(info_of_params_to_include)
f, axarr = plt.subplots(nrows=n, ncols=n, figsize=figure_size)
for i in range(n):
for j in range(n):
# get the current axis
ax = axarr[i, j]
if j == 0:
ax.set_ylabel(info_of_params_to_include[i].label)
if i == n - 1:
ax.set_xlabel(info_of_params_to_include[j].label)
if i == j:
# plot histogram
Fig.add_histogram_to_ax(
ax=ax,
data=info_of_params_to_include[i].values,
x_range=info_of_params_to_include[i].range)
ax.set_yticklabels([])
ax.set_yticks([])
else:
ax.scatter(info_of_params_to_include[j].values,
info_of_params_to_include[i].values,
alpha=0.5,
s=2)
ax.set_xlim(info_of_params_to_include[j].range)
ax.set_ylim(info_of_params_to_include[i].range)
# correlation line
b, m = polyfit(info_of_params_to_include[j].values,
info_of_params_to_include[i].values, 1)
ax.plot(info_of_params_to_include[j].values,
b + m * info_of_params_to_include[j].values,
'-',
c='black')
corr, p = pearsonr(info_of_params_to_include[j].values,
info_of_params_to_include[i].values)
ax.text(0.95,
0.95,
'{0:.2f}'.format(corr),
transform=ax.transAxes,
fontsize=6,
va='top',
ha='right')
f.align_ylabels(axarr[:, 0])
f.tight_layout()
f.savefig(fig_filename, bbox_inches='tight', dpi=300)