Module `ambit_stochastics.helpers.cov_estimator`

Expand source code

import numpy  as np

def compute_empirical_covariance_matrix(values,max_lag_time,max_lag_space):
    """Check the 'Simple Ambit field example usage' Jupyter notebook to see an example.

    Args:
      values: a numpy array containing the values of the simulated ambit field
      max_lag_time: positive integer: the maximum number of lags at which to estimate autocorrelation and autocovariance on the time axis
      max_lag_space: positive integer: the maximum number at which to estimate autocorrelation and autocovaraince on the space axis
          
    Returns:
      result_cov: a numpy array with shape
      result_cor: a numpy array with shape
    """
    nr_simulations, nr_rows, nr_columns = values.shape
    result_cov = np.zeros((nr_simulations,max_lag_time+1,max_lag_space+1))
    result_cor = np.zeros((nr_simulations,max_lag_time+1,max_lag_space+1))

    
    for row in range(max_lag_time+1):
        for column in range(max_lag_space+1):
            
            nr_elements = (nr_rows - row)*(nr_columns - column)
            
            sub_matrix_1 = values[:,:nr_rows - row, :nr_columns - column]
            sub_matrix_2 = values[:,row :, column :]
            #assert sub_matrix_1.shape == sub_matrix_2.shape
            
            mean_1  = np.einsum('ijk->i',sub_matrix_1) / nr_elements
            mean_2  = np.einsum('ijk->i',sub_matrix_2) / nr_elements
            
            variance_estimator_1 = np.array([np.var(sub_matrix_1[i,:,:]) for i in range(nr_simulations)])
            variance_estimator_2 = np.array([np.var(sub_matrix_2[i,:,:]) for i in range(nr_simulations)])

            
            sub_matrix_1 = sub_matrix_1 - mean_1[:,np.newaxis,np.newaxis]
            sub_matrix_2 = sub_matrix_2 - mean_2[:,np.newaxis,np.newaxis]
            
            covariances  = np.einsum('ijk,ijk->i',sub_matrix_1,sub_matrix_2) / nr_elements
            
            result_cov[:,row,column] = covariances
            result_cor[:,row,column] = covariances/(variance_estimator_1 * variance_estimator_2)**0.5
            
    return result_cov,result_cor

Functions

def compute_empirical_covariance_matrix(values, max_lag_time, max_lag_space)

Check the 'Simple Ambit field example usage' Jupyter notebook to see an example.

Args

values: a numpy array containing the values of the simulated ambit field
max_lag_time: positive integer: the maximum number of lags at which to estimate autocorrelation and autocovariance on the time axis
max_lag_space: positive integer: the maximum number at which to estimate autocorrelation and autocovaraince on the space axis

Returns

result_cov: a numpy array with shape
result_cor: a numpy array with shape

Expand source code

def compute_empirical_covariance_matrix(values,max_lag_time,max_lag_space):
    """Check the 'Simple Ambit field example usage' Jupyter notebook to see an example.

    Args:
      values: a numpy array containing the values of the simulated ambit field
      max_lag_time: positive integer: the maximum number of lags at which to estimate autocorrelation and autocovariance on the time axis
      max_lag_space: positive integer: the maximum number at which to estimate autocorrelation and autocovaraince on the space axis
          
    Returns:
      result_cov: a numpy array with shape
      result_cor: a numpy array with shape
    """
    nr_simulations, nr_rows, nr_columns = values.shape
    result_cov = np.zeros((nr_simulations,max_lag_time+1,max_lag_space+1))
    result_cor = np.zeros((nr_simulations,max_lag_time+1,max_lag_space+1))

    
    for row in range(max_lag_time+1):
        for column in range(max_lag_space+1):
            
            nr_elements = (nr_rows - row)*(nr_columns - column)
            
            sub_matrix_1 = values[:,:nr_rows - row, :nr_columns - column]
            sub_matrix_2 = values[:,row :, column :]
            #assert sub_matrix_1.shape == sub_matrix_2.shape
            
            mean_1  = np.einsum('ijk->i',sub_matrix_1) / nr_elements
            mean_2  = np.einsum('ijk->i',sub_matrix_2) / nr_elements
            
            variance_estimator_1 = np.array([np.var(sub_matrix_1[i,:,:]) for i in range(nr_simulations)])
            variance_estimator_2 = np.array([np.var(sub_matrix_2[i,:,:]) for i in range(nr_simulations)])

            
            sub_matrix_1 = sub_matrix_1 - mean_1[:,np.newaxis,np.newaxis]
            sub_matrix_2 = sub_matrix_2 - mean_2[:,np.newaxis,np.newaxis]
            
            covariances  = np.einsum('ijk,ijk->i',sub_matrix_1,sub_matrix_2) / nr_elements
            
            result_cov[:,row,column] = covariances
            result_cor[:,row,column] = covariances/(variance_estimator_1 * variance_estimator_2)**0.5
            
    return result_cov,result_cor