InterventionEffectCalculator: InterventionEffectCalculator

In frbl/OnlineSuperLearner: Online SuperLearner package

Description

Class that can be used to calculate the effect of a specific intervention. For the specification of how to define interventions, look at the InterventionParser class.

Usage

InterventionEffectCalculator

Format

An object of class R6ClassGenerator of length 24.

Methods

initialize(bootstrap_iterations, outcome_variable

Creates a new InterventionEffectCalculator.

@param bootstrap_iterations integer the number of bootstrap iterations to use when calculating the effect of an intervention. If there are more iterations, the result will be more reliable, but calculating will take significantly longer.

@param outcome_variable string the name of the variable for which the outcome should be measured. This is usualy Y (depending on you setup).

@param parallel (default = TRUE) should the effect calculator run in a multithreaded / multicore way?

@param verbose (default = FALSE) the verbosity to use when running the intervention calculation

calculate_intervention_effect(osl, interventions, discrete, initial_data, tau, check = FALSE)

Calculates the actual effect of a set of interventions. This function will call the evaluate_single_intervention for each of the interventions in the list of interventions passed to this functions. When calling this function one can specify whether to use the discrete OSL (discrete = TRUE) or whether to uyse the contionous OSL (discrete = FALSE).

@param osl OnlineSuperLearner object the fitted OnlineSuperLearner instance. This instance is used to perform the predictions with.

@param interventions list a list of interventions to use when calculating the effect of the interventions. Each of these interventions is processed in sequence and the result is stored in a list.

@param discrete boolean should the discrete superlearner (or the cts superlearner) be used?

@param initial_data data.table the first row / set of blocks needed to initialize the intervention estimation algorithm. This block needs to be large enough to provide enough historical data for all summary measures.

@param tau integer the time t at which we'd like to measure the effect of the intervention

@param check (default = FALSE) boolean, should the input parameters be checked for correctness? This might make the process slighly slower.

@return a list with intervention effects for each of the specified interventions. Each intervention entry then contains a vector of intervention effects.

evaluate_single_intervention(osl, initial_data, intervention, tau, discrete)

Calculates the effect of a single intervention. The specification of the intervention needs to comply with the definition in the InterventionParser class. One can specify whether or not to use the discrete super learner by toggling the discrete argument.

@param osl OnlineSuperLearner object the fitted OnlineSuperLearner instance. This instance is used to perform the predictions with.

@param initial_data data.table the first row / set of blocks needed to initialize the intervention estimation algorithm. This block needs to be large enough to provide enough historical data for all summary measures.

@param intervention an intervention to run on the data.

@param tau integer the time t at which we'd like to measure the effect of the intervention

@param discrete boolean should the discrete superlearner (or the cts superlearner) be used?

@return a vector of intervention effects for each iteration of intervention.

perform_initial_estimation(data, intervention, tau)

This function can be used to generate an initial estimation, calculated using the plain OSL. This method then returns a value given the provided data, tau, and intervention. Essentially, this does the same as the calculate_intervention_effect function, but returns the mean of all intervention bootstraps.

@param data.table the data to seed the sampling procedure.

@param intervention the intervention itself, see InterventionParser for more details

@param tau integer the time at which we want to evaluate the intervention

@return double the estimated mean over all bootstrap iterations

is_parallel()

Active method. Function to check whether the current object runs in parallel or not.

@return boolean TRUE if parallel.

get_bootstrap_iterations()

Active method. Returns the number of bootstrap iterations specified.

@return integer the number of bootstrap iterations.

get_relevant_variables)

Active method. The relevant variables specified when intializing the object.

@return list of RelevantVariable objects.

get_outcome_variable()

Active method. Returns the outcome variable specified on initialization (string).

@return string the string representation of the name of the outcome variable.

frbl/OnlineSuperLearner documentation built on Feb. 9, 2020, 9:28 p.m.