environmentaltools.temporal.fit_marginal_distribution

environmentaltools.temporal.fit_marginal_distribution(df: DataFrame, parameters: dict, verbose: bool = False)[source]

Fits a stationary (or not), simple or mixed probability model to data.

Additional information can be found in Cobos et al., 2022, ‘MarineTools.temporal: A Python package to simulate Earth and environmental time series’. Environmental Modelling and Software.

Parameters:

df (pd.DataFrame) – The raw time series
parameters (dict) –
The initial guess parameters of the probability models with the following keys:
- ’var’str
  Name of the variable
- ’type’str
  Defines circular or linear variables
- ’fun’list
  List of strings with the name of the probability model
- ’non_stat_analysis’bool
  False for stationary, True for non-stationary
- ’ws_ps’float or list
  Initial guess of percentiles or weights of PMs
- ’basis_function’dict or None
  GFS expansion specification:
  - ’method’str
    Option for the GFS
  - ’no_terms’int
    Number of terms of GFS
  - ’periods’list
    Periods of oscillation for NS-PMs
- ’transform’dict or None
  Normalization options:
  - ’make’bool
    Whether to apply transformation
  - ’method’str
    ’box-cox’ or ‘yeo-johnson’
  - ’plot’bool
    Whether to plot
- ’detrend’dict or None
  Removing trends options:
  - ’make’bool
    Whether to detrend
  - ’method’str
    GFS option (commonly polynomial approaches)
- ’optimization’dict
  Parameters for scipy.optimize.minimize:
  - ’method’str
    e.g., “SLSQP”
  - ’maxiter’float
    Maximum iterations
  - ’ftol’float
    Function tolerance
  - ’eps’float
    Step size for numerical derivatives
  - ’bounds’float
    Bounds for optimization
  - ’weighted’bool
    For weighted data along time axis
- ’giter’int
  Number of global iterations
- ’initial_parameters’dict or None
  Initial parameters for unique optimization mode:
  - ’make’bool
    Whether to use initial parameters
  - ’mode’list
    Mode to be computed independently
  - ’par’list
    Initial guess of parameters for given mode
- ’file_name’str, optional
  Path where analysis will be saved
verbose (bool, optional) – If True, shows information of the fitting process. Default is False.

Returns:

The fitting parameters

Return type:

dict

Examples

>>> param = {'Hs': {'var': 'Hs',
...                 'fun': {0: 'norm'},
...                 'type': 'linear',
...                 'non_stat_analysis': True,
...                 'basis_function': {"method": "trigonometric",
...                                   "no_terms": 4,
...                                   "periods": [1, 2, 4]},
...                 'ws_ps': 1,
...                 'transform': {"make": True,
...                              "plot": False,
...                              "method": "box-cox"},
...                 'detrend': {"make": True,
...                            "method": "polynomial"},
...                 'optimization': {'method': 'SLSQP',
...                                 'eps': 1e-7,
...                                 'ftol': 1e-4,
...                                 'maxiter': 1e2,
...                                 'bounds': 0.5},
...                 'giter': 10,
...                 'scale': False,
...                 'bic': True,
...                 'file_name': 'output.pkl'
...                 }
...         }