BinaryTargetData class

A class for binary target data objects.

Super class

TargetData -> BinaryTargetData

Public fields

control_rate

Rate in the control arm of the target study

treatment_rate

Rate in the treatment arm of the target study

Methods

Public methods

• BinaryTargetData$new()

• BinaryTargetData$generate()

• BinaryTargetData$samples_from_counts()

• BinaryTargetData$enumerate_support()

• BinaryTargetData$to_dict()

• BinaryTargetData$clone()

Inherited methods

• TargetData$plot_sample()

---

BinaryTargetData$new()

Initializes the binary target data object.

Usage

BinaryTargetData$new(
  source_data,
  sampling_approximation,
  target_sample_size_per_arm,
  control_drift,
  treatment_drift,
  summary_measure_likelihood
)

Arguments

source_data

The source data object.

sampling_approximation

The sampling approximation flag.

target_sample_size_per_arm

The target sample size per arm.

control_drift

The control drift.

treatment_drift

The treatment drift.

summary_measure_likelihood

The summary measure distribution.

---

BinaryTargetData$generate()

Generates samples for the binary target data object.

Usage

BinaryTargetData$generate(n_replicates)

Arguments

n_replicates

The number of replicates to generate.

Returns

A data frame containing the generated samples.

---

BinaryTargetData$samples_from_counts()

The replicate rows a trial with the given responder counts is analysed from. generate() draws the counts and enumerate_support() lists them, and both build their rows here, so the two cannot disagree on what a trial looks like to the analysis.

Usage

BinaryTargetData$samples_from_counts(
  n_control_responders,
  n_treatment_responders
)

Arguments

n_control_responders

Integer vector of control-arm responders.

n_treatment_responders

Integer vector of treatment-arm responders, the same length.

Returns

A data frame with one row per pair of counts.

---

BinaryTargetData$enumerate_support()

Every trial outcome with non-negligible probability, with its probability, so that an operating characteristic can be computed as an exact weighted sum over trials instead of a Monte Carlo average. Each arm's responder count is kept between its tail_mass / 4 and 1 - tail_mass / 4 binomial quantiles, so at most tail_mass is left out across both arms, and the weights, the product of the two binomial probabilities, are renormalised over the pairs kept.

Usage

BinaryTargetData$enumerate_support(tail_mass = 1e-10)

Arguments

tail_mass

Upper bound on the probability of the trials left out.

Returns

A list: samples, the replicate rows as generate() builds them; weights, their probabilities, summing to 1; and omitted_mass, the probability of the trials left out.

---

BinaryTargetData$to_dict()

Converts the target data object to a dictionary.

Usage

BinaryTargetData$to_dict()

Returns

A list representing the target data object.

---

BinaryTargetData$clone()

The objects of this class are cloneable with this method.

Usage

BinaryTargetData$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.