Type: Package
Title: Bayesian Response-Adaptive Design Analysis
Version: 1.0
Date: 2023-01-18
Description: Provides access to a range of functions for analyzing, applying and visualizing Bayesian response-adaptive trial designs for a binary endpoint. Includes the predictive probability approach and the predictive evidence value designs for binary endpoints.
Imports: methods, fbst, extraDistr, doParallel, foreach, parallel, doSNOW, progress, cli
Suggests: knitr, rmarkdown, DT
License: GPL-3
VignetteBuilder: knitr
NeedsCompilation: no
Packaged: 2023-01-23 14:46:54 UTC; riko
Author: Riko Kelter ORCID iD [aut, cre]
Maintainer: Riko Kelter <riko.kelter@uni-siegen.de>
Repository: CRAN
Date/Publication: 2023-01-24 10:40:11 UTC

Bayesian Response-Adaptive Design Analysis

Description

Provides access to a range of functions for analyzing, applying and visualizing Bayesian response-adaptive trial designs for a binary endpoint. Includes the predictive probability approach and the predictive evidence value designs for binary endpoints.

Details

Provides access to a range of functions for analyzing, applying and visualizing
Bayesian response-adaptive trial designs for a binary endpoint. Includes the
predictive probability approach and the predictive evidence value designs for
binary endpoints.

Package: brada
Type: Package
Title: Bayesian Response-Adaptive Design Analysis
Version: 1.0
Date: 2023-01-18
Authors@R: c(person(given = "Riko", family = "Kelter", role = c("aut", "cre"), email = "riko.kelter@uni-siegen.de", comment = c(ORCID = "0000-0001-9068-5696")))
Description: Provides access to a range of functions for analyzing, applying and visualizing Bayesian response-adaptive trial designs for a binary endpoint. Includes the predictive probability approach and the predictive evidence value designs for binary endpoints.
Imports: methods, fbst, extraDistr, doParallel, foreach, parallel, doSNOW, progress, cli
Suggests: knitr, rmarkdown, DT
License: GPL-3
VignetteBuilder: knitr
Author: Riko Kelter [aut, cre] (<https://orcid.org/0000-0001-9068-5696>)
Maintainer: Riko Kelter <riko.kelter@uni-siegen.de>

Index of help topics: 

$,brada-method          Returns an object from an object of class
                        'brada'.
brada                   brada
brada-class             Class '"brada-class"'
brada-package           Bayesian Response-Adaptive Design Analysis
calibrate               calibrate
generateData            generateData
monitor                 monitor
names.brada             names.brada
plot.brada              plot.brada
power                   power
show.brada              show.brada
summary.brada           summary.brada

Author(s)

NA

Maintainer: NA

Returns an object from an object of class brada.

Description

Returns an object from an object of class brada

Details

-

Value

No return value.

Author(s)

Riko Kelter

brada

Description

Performs a Bayesian response-adaptive design analysis for trials with a binary endpoint.

Usage

brada(a0=1,b0=1,Nmax=40,batchsize=5,nInit,p_true,p0,p1,
theta_T=0.90,theta_L=0.1,theta_U=1,nsim=100,
seed=42,method="PP",refFunc="flat",nu=0,
shape1=1,shape2=1,truncation=1,cores=2)

Arguments

a0

shape1 parameter of the beta prior.

b0

shape2 parameter of the beta prior.

Nmax

Maximum trial size.

batchsize

sample size after which an interim analysis is performed.

nInit

Initial sample size at which the first interim analysis is performed.

p_true

True binary response probability used for simulation.

p0

Right boundary of the null hypothesis to be tested.

p1

Left boundary of the alternative hypothesis to be tested.

theta_T

Threshold used in the designs for including trajectories as evidential.

theta_L

Stopping threshold for futility.

theta_U

Stopping threshold for efficacy.

nsim

Number of Monte Carlo iterations.

seed

Random number generator seed.

cores

Number of CPU cores to be used for computation. Defaults to 2, but 4 or larger is recommended.

method

Can be either "PP" or "PPe", depending on whether the predictive probability approach or the predictive evidence value design is desired. Note that the former is a special case of the latter.

refFunc

A string, either "flat", "beta", "binaryStep", "relu", "palu" or "lolu". See vignettes for explanation.

nu

A numeric value larger or equal to zero, indicating which evidence threshold if used in the predictive evidence value design.

shape1

shape1 parameter of the beta reference function, if used.

shape2

shape2 parameter of the beta reference function, if used.

truncation

Truncation point in case an artificial neural network reference function is used.

Value

Returns an object of class brada.

Author(s)

Riko Kelter

Examples

pp_design = brada(Nmax = 30, batchsize = 5, nInit = 10, 
               p_true = 0.2 , p0 = 0.2, p1 = 0.2, 
               nsim = 10,
               a0 = 1, b0 = 1, 
               theta_T = 0.90, theta_L = 0.1, theta_U = 1, 
               method = "PP",
               cores = 2)
summary(pp_design)

Class "brada-class"

Description

Class for modelling the results of a Bayesian response-adaptive design analysis

Objects from the Class

Store the results of a Bayesian response-adaptive design analysis

Slots

data:

Object of class "list" holding the results of the Bayesian response-adaptive design analysis. a0 and b0 store the beta prior shape parameters, Nmax and batchsize store the maximum trial size and the batchsize used for interim analyses. nInit is the minimum sample size at which the first interim analysis is conducted. p_true is the true response probability used for simulation, p0 is the right boundary of the null hypothesis and p1 the left boundary of the alternative hypothesis. ...

calibrate

Description

Calibrates a brada object to achieve specified false-positive and false-negative rates.

Usage

calibrate(brada_object, nsim = 100, cores = 2, seq, 
alpha=NULL, beta=NULL, calibration = "nu")

Arguments

brada_object

An object of class brada

nsim

Number of Monte Carlo iterations

cores

Number of cores used for computation

seq

Sequence of values for the evidence threshold nu, or sequence of the futility threshold theta_L, depending on which value is passed to the function in the calibration argument. For example, seq=seq(0,1,0.1) in combination with calibration="nu" and alpha=0.1 implies that the function tries to calibrate the false-positive rate to decrease below alpha=0.1 by iteratively increasing nu from zero to one in steps of size 0.1.

alpha

Upper bound for false-positive rate. Note that it is only possible to specify either alpha or beta. When alpha is passed as an argument, calibration should take the value "nu", while when beta is passed as an argument, calibration should take the value "theta_L".

beta

Upper bound for false-negative rate

calibration

String which specifies which parameter to calibrate. Can take the values "nu" or "theta_L".

Value

Prints the output to the console and returns the false-positive rate or false-negative rate of the calibrated design, depending on which value the calibration argument takes.

Author(s)

Riko Kelter

generateData

Description

Generates a matrix of trial data.

Usage

generateData(p,Nmax,nsim,seed=420)

Arguments

p

true response probability

Nmax

Maximum trial size.

nsim

Number of Monte Carlo iterations.

seed

Random number generator seed.

Value

Returns a matrix with simulated trial data.

Author(s)

Riko Kelter

Examples

generateData(p=0.2,Nmax=40,nsim=100,seed=420)

monitor

Description

Monitors a running trial with a binary endpoint and calculates the predictive probability or predictive evidence that the trial will result in a success. Reports whether to stop early for futility or efficacy based on a vector of binary observations.

Usage

monitor(brada_object, obs)

Arguments

brada_object

An object of class brada.

obs

A vector of binary observations, where 1 is a success (response) and 0 a failure (no response).

Value

No return value, prints the result of the monitoring to the console.

Author(s)

Riko Kelter

Examples

design = brada(Nmax = 40, batchsize = 5, nInit = 10, 
                   p_true = 0.2 , p0 = 0.2, p1 = 0.2, 
                   nsim = 100,
                   a0 = 1, b0 = 1, 
                   theta_T = 0.95, theta_L = 0.05, theta_U = 0.975, 
                   method = "PP",
                   cores = 2)
monitor(design, obs = c(0,1,1,0,0,1,0,1,1,1))

names.brada

Description

Plots the names of the objects stored in the brada object of a Bayesian response-adaptive design analysis.

Usage

## S3 method for class 'brada'
names(x)

Arguments

x

An Object of class "brada".

Details

Plots the names of the objects stored in the trials object of a Bayesian response-adaptive design analysis.

Value

Returns a list of names.

Author(s)

Riko Kelter

plot.brada

Description

Plots the results of a Bayesian response-adaptive design analysis.

Usage

## S3 method for class 'brada'
plot(x, trajectories = 100, ...)

Arguments

x

An Object of class "brada".

trajectories

Number of trajectories to be plotted. Defaults to 100.

...

Additional parameters, see "plot(x, ...)".

Value

Returns a plot.

Author(s)

Riko Kelter

Examples

design = brada(Nmax = 40, batchsize = 5, nInit = 10, 
               p_true = 0.2 , p0 = 0.2, p1 = 0.2, 
               nsim = 100,
               a0 = 1, b0 = 1, 
               theta_T = 0.90, theta_L = 0.1, theta_U = 1, 
               method = "PP",
               cores = 2)
plot(design)

power

Description

Performs a power analysis for a brada object.

Usage

power(brada_object, p_true, nsim=100, cores=2)

Arguments

brada_object

An object of class brada

p_true

the true response probability used for the power analysis

nsim

the number of Monte Carlo simulation, defaults to 100.

cores

CPU cores used for computation. Defaults to 2.

Value

Returns an object of class brada.

Author(s)

Riko Kelter

Examples

design = brada(Nmax = 30, batchsize = 5, nInit = 10, 
               p_true = 0.2 , p0 = 0.2, p1 = 0.2, 
               nsim = 1000,
               a0 = 1, b0 = 1, 
               theta_T = 0.90, theta_L = 0.1, theta_U = 1, 
               method = "PP",
               cores = 1)
    design_power = power(design, p_true = 0.4, nsim = 1000)
    plot(design_power)

show.brada

Description

Prints the main results of a Bayesian response-adaptive design analysis to the console.

Usage

## S3 method for class 'brada'
show(object)

Arguments

object

An Object of class "brada".

Details

Shows the main results of a Bayesian response-adaptive design analysis stored in an object of class brada.

Value

Prints the results onto the console.

Author(s)

Riko Kelter

summary.brada

Description

Prints the results of a Bayesian response-adaptive design analysis.

Usage

## S3 method for class 'brada'
summary(object, ...)

Arguments

object

An Object of class "brada".

...

Additional parameters, see "summary(object, ...)".

Details

Summarises the results of a Bayesian response-adaptive design analysis.

Value

Prints the results onto the console.

Author(s)

Riko Kelter

Examples

pp_design = brada(Nmax = 40, batchsize = 5, nInit = 10, 
               p_true = 0.2 , p0 = 0.2, p1 = 0.2, 
               nsim = 100,
               a0 = 1, b0 = 1, 
               theta_T = 0.90, theta_L = 0.1, theta_U = 1, 
               method = "PP",
               cores = 2)
summary(pp_design)