This class inherits from GaussianEmpiricalBayesPP and implements the PDCCPP method.
Super classes
Model -> GaussianConjugate -> GaussianStaticBorrowing -> GaussianEmpiricalBayesPP -> GaussianPDCCPP
Methods
Inherited methods
Model$calibrate_for_design()Model$check_data()Model$create()Model$estimate_bayesian_operating_characteristics()Model$estimate_frequentist_operating_characteristics()Model$hypothesis_space_transformation()Model$inference_cache_scope()Model$plot_pdfs()Model$plot_posterior_pdf()Model$plot_prior_pdf()Model$posterior_beta_mixture()Model$posterior_ess()Model$posterior_quantile()Model$print_model_summary()Model$prior_elir_ess()Model$simulation_for_given_treatment_effect()Model$test_decision()GaussianConjugate$credible_interval()GaussianConjugate$posterior_cdf()GaussianConjugate$posterior_mean()GaussianConjugate$posterior_median()GaussianConjugate$posterior_moments()GaussianConjugate$posterior_pdf()GaussianConjugate$posterior_to_RBesT()GaussianConjugate$posterior_variance()GaussianConjugate$prior_cdf()GaussianConjugate$prior_to_RBesT()GaussianConjugate$sample_posterior()GaussianConjugate$sample_prior()GaussianConjugate$vectorised_replicate_inference()GaussianStaticBorrowing$summary_rows()GaussianEmpiricalBayesPP$empirical_bayes_update()GaussianEmpiricalBayesPP$inference()GaussianEmpiricalBayesPP$plot_power_parameter_vs_drift()GaussianEmpiricalBayesPP$prior_pdf()GaussianEmpiricalBayesPP$vectorised_posterior_parameters()GaussianEmpiricalBayesPP$vectorised_prior_variance()
GaussianPDCCPP$new()
Initialize the GaussianPDCCPP object.
Usage
GaussianPDCCPP$new(prior, theta_0, null_space)GaussianPDCCPP$power_parameter_estimation()
GaussianPDCCPP$vectorised_power_parameter()
The power parameter for every replicate at once
The calibration depends on a replicate only through its target sampling
variance, and it is a smooth function of it. Rather than search for it
once per replicate, it is computed at 200 variances spaced evenly on the
log scale across the replicates' range and interpolated linearly. Near
the borrowing cut-off the power parameter is very sensitive to the
calibration, so these few searches are run to a tolerance of 1e-9
rather than the configured one: the interpolated calibration is then
closer to the exact one than a per-replicate search at the configured
tolerance would be. Equation (9) is evaluated for every replicate,
exactly as power_parameter_estimation() does.
GaussianPDCCPP$calibration_parameter()
The calibration parameter z_1-c/2 for one target sampling variance
Usage
GaussianPDCCPP$calibration_parameter(
target_data_sampling_variance,
target_sample_size_per_arm,
source_treatment_effect_estimate,
tolerance = self$parameters$tolerance
)GaussianPDCCPP$power_parameter_from_calibration()
The power parameter given the calibration, equation (9) of Nikolakopoulos et al (2018)
Vectorised over its arguments, so it serves one replicate or all of them.
Usage
GaussianPDCCPP$power_parameter_from_calibration(
target_treatment_effect_estimate,
source_treatment_effect_estimate,
target_data_sampling_variance,
target_sample_size_per_arm,
calibration_parameter
)Arguments
target_treatment_effect_estimateTarget estimates, after
hypothesis_space_transformation().source_treatment_effect_estimateSource estimate, after the same transformation.
target_data_sampling_varianceTarget sampling variances, per patient.
target_sample_size_per_armTarget sample size per arm.
calibration_parameterCalibration parameters z_1-c/2.