17,000+ users
Click for interactive map
 File NameSizeNotes
 CATBUBDesign.zip  511 KB Contains R code, examples, documents. (This program was formerly called "BUB Design")
Email Address: Organization: Occupation:
Your email will be used to send you occasional information about software updates. It will NOT be shared with third parties.

Categorical Outcome Bayesian Utility Based (CATBUB) Designs

CATBUB designs use numerical utilities as a one-dimensional comparator for treatments whose performance is determined by categorical outcomes, e.g., trinary, bivariate-binary, ordinal, etc.

Description

The software, i.e. an R program called "CATBUB_Design.R," contains design and analysis functions catbub.design() and catbub.analysis() that facilitate the implementation of a group sequential CATBUB design. This file is extensively annotated and explains the inputs and outputs for every function defined within. Briefly, the design function implements the computational CATBUB design algorithms of Murray, Thall and Yuan [1]. The investigator supplies the targeted alternative and design parameters, i.e. alpha, beta, rho and planned interim look sampling fractions, and the function returns a group sequential design, i.e. planned maximum sample size, probability thresholds, and operating characteristics at the supplied alternatives. The analysis function allows the investigator to supply the observed data, planned maximum sample size and previous interim sample sizes, and returns the decision, i.e. "Continue Enrollment", "A>B" or "B>A."

Features

  • Written in R
  • Computationally Efficient
  • Easy to use
  • Well documented

Illustrations

We also include illustrative R programs for trinary, bivariate-binary, bivariate-ordinal, and the CLL outcome. We implement fixed sample, i.e. a single analysis, and group sequential designs for each outcome type. For the trinary and bivariate-binary outcomes, we provide an example program for assessing sensitivity to the elicited utilities. For the trinary outcome, we provide an example of how to identify the relevant targeted alternative among a set of targeted alternatives to use during the design.

System Requirements

  • R v2.12.1 or newer, freely-available on CRAN

Credits

Thomas A. Murray implemented the numerical algorithms and authored the illustrative examples.

References

[1] Murray, Thomas A., Thall, Peter F., and Yuan, Ying (2016) "Utility-Based Designs for Randomized Comparative Trials with Categorical Outcomes," Accepted to Statistics in Medicine

[2] Murray, Thomas A., Thall, Peter F., and Yuan, Ying (2016) "Supplementary Materials: Utility-Based Designs for Randomized Comparative Trials with Categorical Outcomes," Accepted to Statistics in Medicine