def __init__(self,
basename=None,
parameter_data=None,
res_df=None,
jco_df=None):
''' Create ObSen class
Parameters
----------
basename : str, optional
basename for PEST control file, if full path not provided the
current working directory is assumed. Optional but must be provided
if any of parameter_data, res_df, or jco_df are not provided.
parameter_data : DataFrame, optional
Pandas DataFrame of the paramter data from a .pst file. If not
provided it will be read in based on the base name of pest file
provided.
jco_df : DataFrame, optional
Pandas DataFrame of the jacobian. If not provided then it will be
read in based on base name of pest file provided. Providing a
jco_df offers some efficiencies if working interactively.
Otherwise the jco is read in every time ObSen class is initialized.
res_df : DataFrame, optional
Residual DataFrame used to define the weights to
calculate the observation sensitivity. Providing a
res_df offers some efficiencies if working interactively.
If not provided it will look for basename+'.res'.
Weights are not taken from PEST control file
(.pst) because regularization weights in PEST conrtrol file do
not reflect the current weights.
Attributes
----------
df : Pandas DataFrame
DataFrame of observation sensitivity. Index entries of the DataFrame
are the observation names.
Methods
-------
#plot()
tail()
head()
#par()
group()
sum_group()
#plot_sum_group()
#plot_mean_group()
Notes
------
'''
if basename is not None:
self.basename = os.path.split(basename)[-1].split('.')[0]
self.directory = os.path.split(basename)[0]
if len(self.directory) == 0:
self.directory = os.getcwd()
if jco_df is None:
jco_file = os.path.join(self.directory, self.basename + '.jco')
jco = Jco()
jco.from_binary(jco_file)
self.jco_df = jco.to_dataframe()
else:
self.jco_df = jco_df
if res_df is None:
res_file = os.path.join(self.directory, self.basename + '.res')
pst = Pst(filename=None, load=False, resfile=res_file)
self.res_df = pst.load_resfile(res_file)
else:
self.res_df = res_df
# Set index of res_df
self.res_df.set_index('name', drop=False, inplace=True)
# Build _obs_data
weights = []
ob_groups = []
obs = []
for ob in self.jco_df.index:
weight = self.res_df.loc[ob.lower()]['weight']
ob_group = self.res_df.loc[ob.lower()]['group']
weights.append(weight)
ob_groups.append(ob_group)
obs.append(ob)
self._obs_data = pd.DataFrame({
'OBSNME': obs,
'OBGNME': ob_groups,
'WEIGHT': weights,
'ObSen_Weight': weights
})
self._obs_data.set_index('OBSNME', inplace=True)
# Fill DataFrame
self.df = self.calc_sensitivity()
def __init__(self, basename=None, parameter_data=None, res_df=None,
jco_df=None, drop_regul=False, drop_groups=None,
keep_groups=None, keep_obs=None, remove_obs=None):
''' Create ParSen class
Parameters
----------
basename : str, optional
basename for PEST control file, if full path not provided the
current working directory is assumed. Optional but must be provided
if any of parameter_data, res_df, or jco_df are not provided.
parameter_data : DataFrame, optional
Pandas DataFrame of the paramter data from a .pst file. If not
provided it will be read in based on the base name of pest file
provided.
jco_df : DataFrame, optional
Pandas DataFrame of the jacobian. If not provided then it will be
read in based on base name of pest file provided. Providing a
jco_df offers some efficiencies if working interactively.
Otherwise the jco is read in every time ParSen class is initialized.
res_df : DataFrame, optional
Residual DataFrame used to define the weights to
calculate the parameter sensitivity. Providing a
res_df offers some efficiencies if working interactively.
If not provided it will look for basename+'.res'.
Weights are not taken from PEST control file
(.pst) because regularization weights in PEST conrtrol file do
not reflect the current weights.
drop_regul: {False, True}, optional
Flag to drop regularization information in calculating parameter
sensitivity. Will set weight to zero for all observations with
'regul' in the observation group name
drop_groups: list, optional
List of observation groups to drop when calculating parameter
sensitivity. If all groups are part of regularization it may
be easier to use the drop_regul flag
keep_groups: list, optional
List of observation groups to include in calculating parameter
sensitivity. Sometimes easier to use when looking at sensitivity
to a single, or small number, or observation groups
keep_obs: list, optional
List of observations to include in calculating parameter
sensitivity. If keep_obs != None then weights for all observations
not in keep_obs will be set to zero.
remove_obs: list, optional
List of observations to remove in calculating parameter
sensitivity. If remove_obs != None then weights for all
observations in remove_obs will be set to zero.
Attributes
----------
df : Pandas DataFrame
DataFrame of parameter sensitivity. Index entries of the DataFrame
are the parameter names. The DataFrame has two columns:
1) Parameter Group and 2) Sensitivity
Methods
-------
plot()
tail()
head()
par()
group()
sum_group()
plot_sum_group()
plot_mean_group()
Notes
------
'''
if basename is not None:
self.basename = os.path.split(basename)[-1].split('.')[0]
self.directory = os.path.split(basename)[0]
if len(self.directory) == 0:
self.directory = os.getcwd()
if jco_df is None:
jco_file = os.path.join(self.directory, self.basename + '.jco')
jco = Jco()
jco.from_binary(jco_file)
self.jco_df = jco.to_dataframe()
else:
self.jco_df = jco_df
if res_df is None:
res_file = os.path.join(self.directory, self.basename + '.res')
pst = Pst(filename=None, load=False, resfile=res_file)
self.res_df = pst.load_resfile(res_file)
else:
self.res_df = res_df
# Set index of res_df
self.res_df.set_index('name', drop=False, inplace = True)
if parameter_data is None:
pst_file = os.path.join(self.directory, self.basename + '.pst')
pst = Pst(pst_file, load=True, resfile = None)
self.parameter_data = pst.parameter_data
else:
self.parameter_data = parameter_data
# Build pars_dict
# key is PARNME value is PARGP
self._pars_dict = {}
for index, row in self.parameter_data.iterrows():
self._pars_dict[row['parnme'].lower()] = row['pargp'].lower()
# Build _obs_data
weights = []
ob_groups = []
obs = []
for ob in self.jco_df.index:
weight = self.res_df.loc[ob.lower()]['weight']
ob_group = self.res_df.loc[ob.lower()]['group']
weights.append(weight)
ob_groups.append(ob_group)
obs.append(ob)
self._obs_data = pd.DataFrame({'OBSNME': obs, 'OBGNME': ob_groups, 'WEIGHT': weights, 'ParSen_Weight' : weights})
self._obs_data.set_index('OBSNME', inplace=True)
if drop_regul is True:
self.drop_regul(calc_sensitivity=False)
if drop_groups is not None:
self.drop_groups(drop_groups=drop_groups, calc_sensitivity=False)
if keep_groups is not None:
self.keep_groups(keep_groups = keep_groups, calc_sensitivity=False)
if keep_obs is not None:
self.keep_obs(keep_obs=keep_obs, calc_sensitivity=False)
if remove_obs is not None:
self.remove_obs(remove_obs=remove_obs, calc_sensitivity=False)
# Fill DataFrame
self.df = self.calc_sensitivity()
def __init__(self, basename=None, parameter_data=None, res_df=None,
jco_df=None):
''' Create ObSen class
Parameters
----------
basename : str, optional
basename for PEST control file, if full path not provided the
current working directory is assumed. Optional but must be provided
if any of parameter_data, res_df, or jco_df are not provided.
parameter_data : DataFrame, optional
Pandas DataFrame of the paramter data from a .pst file. If not
provided it will be read in based on the base name of pest file
provided.
jco_df : DataFrame, optional
Pandas DataFrame of the jacobian. If not provided then it will be
read in based on base name of pest file provided. Providing a
jco_df offers some efficiencies if working interactively.
Otherwise the jco is read in every time ObSen class is initialized.
res_df : DataFrame, optional
Residual DataFrame used to define the weights to
calculate the observation sensitivity. Providing a
res_df offers some efficiencies if working interactively.
If not provided it will look for basename+'.res'.
Weights are not taken from PEST control file
(.pst) because regularization weights in PEST conrtrol file do
not reflect the current weights.
Attributes
----------
df : Pandas DataFrame
DataFrame of observation sensitivity. Index entries of the DataFrame
are the observation names.
Methods
-------
#plot()
tail()
head()
#par()
group()
sum_group()
#plot_sum_group()
#plot_mean_group()
Notes
------
'''
if basename is not None:
self.basename = os.path.split(basename)[-1].split('.')[0]
self.directory = os.path.split(basename)[0]
if len(self.directory) == 0:
self.directory = os.getcwd()
if jco_df is None:
jco_file = os.path.join(self.directory, self.basename + '.jco')
jco = Jco()
jco.from_binary(jco_file)
self.jco_df = jco.to_dataframe()
else:
self.jco_df = jco_df
if res_df is None:
res_file = os.path.join(self.directory, self.basename + '.res')
pst = Pst(filename=None, load=False, resfile=res_file)
self.res_df = pst.load_resfile(res_file)
else:
self.res_df = res_df
# Set index of res_df
self.res_df.set_index('name', drop=False, inplace = True)
# Build _obs_data
weights = []
ob_groups = []
obs = []
for ob in self.jco_df.index:
weight = self.res_df.loc[ob.lower()]['weight']
ob_group = self.res_df.loc[ob.lower()]['group']
weights.append(weight)
ob_groups.append(ob_group)
obs.append(ob)
self._obs_data = pd.DataFrame({'OBSNME': obs, 'OBGNME': ob_groups, 'WEIGHT': weights, 'ObSen_Weight' : weights})
self._obs_data.set_index('OBSNME', inplace=True)
# Fill DataFrame
self.df = self.calc_sensitivity()