Predicts the value of a dependent variable in a Quantile Regression model based on the value of the independent variables.
The node accepts the new data to be used when making the prediction on its 'data' input pin. It accepts the file path to the file that contains an R serialized quantile regression model object. The file path can be specified as a Literal value or be obtained from a specified field on the node's optional second input pin.
When run, the node uses the embedded R engine to predict the value for the model's dependent (response) variable based on the values of the independent variables that are present on the node's data input pin.
The names of the fields in the input data must correspond to the names of the independent variables that were used to construct the quantile regression model.
The Summary pin contains a summary of the input serialized model and the file path to the model. The model summary includes information on:
- The call used to generate the model.
- The quantile level.
- The estimates for the coefficients of the independent variables used in the model and the estimate of the intercept.
The predicted value for the dependent variable, together with the corresponding (predictor) values for the independent variables are output on the Results pin.
Where the quantile regression model contains a single quantile level, the node can be configured to show the confidence bound values in the generated output.
The confidence bound value can be set as a value between 0 and 1. By default, this is set to 0.95 (95%).
The node can be configured to specify the confidence method to be used when calculating the confidence intervals. Confidence intervals are only calculated for quantile regression models that contain a single quantile level. One of the following confidence methods can be selected (the default is "Covariance Matrix"):
- Covariance Matrix - Uses an estimated covariance matrix for the property estimates.
- Direct - Uses the "Portnoy and Zhou (1998)" method.
- Percentile - Uses a bootstrap method to generate confidence intervals.
When the confidence interval method is set to "Percentile" a bootstrap method is used to generate confidence intervals. One of the following bootstrap methods can be specified (default is "xy-pair"):
- xy-pair - Use the bootstrap xy-pair method.
- Parzen - Use the bootstrap Parzen method.
- MC - Use the bootstrap MC method.
If the new data contains any missing (NULL) values, the node can be configured to perform one of the following actions (default is "Pass"):
- Pass - Continue processing. A NULL is output in place of the predicted value.
- Fail - Generate an error.
- Omit - Continue processing. A record is not output if the input data has a missing value.
Powered by TIBCO®
Properties
ModelFilePath
Specify the filepath of the Quantile Regression model to be used when predicting dependent values.
Choose the (from Field) variant of this property to look up the value from an input field with the name specified. A value is required for this property.
ShowConfidenceBounds
Optionally specify whether confidence bound values are to be included in the output data.
Confidence bounds are only generated if the specified Quantile Regression model was constructed using only a single quantile estimate. This property is ignored if the model has multiple quantile estimate values. The default value is False.
ConfidenceBounds
Optionally specify the confidence interval bounds for predictions. A positive number between 0 and 1. The default value is 0.95, corresponding with the 95th percentile.
ConfidenceMethod
Optionally specify the method to use when calculating confidence intervals. Choose from:
- Covariance Matrix - Uses an estimated covariance matrix for the property estimates
- Direct - Uses the 'Portnoy and Zhou (1998)' method
- Percentile - Uses a bootstrap method to generate confidence intervals
The default value is Covariance Matrix.
The generation of confidence intervals is only supported for quantile regression models that were constructed with a single quantile estimate (tau), i.e. when the specified model has a 'class' of 'rq'.
Confidence intervals are not calculated for quantile regression models that were constructed with multiple quantile estimates (tau values), i.e. when the specified model has a 'class' of 'rqs'.
When the confidence interval method is set to Percentile the BootstrapMethod property determines the bootstrap method to use when generating the percentile intervals for each prediction.
BootstrapMethod
Optionally specify the bootstrap method to use when calculating the confidence intervals when the ConfidenceMethod property is set to Percentile. Choose from:
- xy-pair - Use the bootstrap xy-pair method
- Parzen - Use the bootstrap Parzen method
- MC - Use the bootstrap MC method
The default value is xy-pair.
MissingValueAction
Optionally specify the action to take if the new data has missing values. Choose from:
- Pass - Continue processing. A NULL is output in place of the predicted value.
- Fail - Generate an error.
- Omit - Continue processing. A record is not output if the input data has a missing value.
The default value is Pass.
Inputs and outputs
Inputs: data, 1 optional.
Outputs: Summary, Results.