Child Class RegressionModelWithInitialResiduals

This class impacts only the method run() (it doesn't change the behaviour of the estimate() method). It computes initial errors of the observations to the predictions (residuals) if run for the first time. The error values are added to dataset as a primary attribute (with the same name as the outcome attribute adding the prefix '_init_error_'). Then in any run, the error is added to the outcome. Thus, in order to compute the residuals, the outcome attribute must be a known attribute of dataset prior to running the model. Note that running this model on the same set of observations which were used for the estimation, should result in the same outcome as the original values of the outcome attribute.

In addition to all parents arguments, this class requires the argument outcome_attribute to be passed into the constructor. The following entries of the run_config dictionary are accepted:

• 'exclude_missing_values_from_initial_error' - default is False.
• 'outcome_attribute_missing_value' - if the above entry is True, this value determines a missing value. Default is 0.
• 'exclude_outliers_from_initial_error' - default is False.
• 'outlier_is_less_than' - lower bound for excluding outliers from computing residuals.
• 'outlier_is_greater_than' - upper bound for excluding outliers from computing residuals.