Computer programs for group sequential tests

Permission to use, copy, modify, and distribute this software and its documentation for any purpose, without fee, and without a written agreement is hereby granted, provided that the above copyright notice appears in all copies.

This code is released on an "as is" basis. There is no warranty. Users are warned that the developers are not liable for any errors or damages that occur as a result of using this code.

If you find these programs of help in your research we would ask that you acknowledge their source in reports and publications.

Programs presented here perform calculations to construct and evaluate group sequential tests. They are provided as a service to readers of Group Sequential Methods with Applications to Clinical Trials by Christopher Jennison and Bruce W Turnbull.

The programs are presented using the notation of the book and, in particular, the method of computation for normally distributed test statistics is as described in our Chapter 19.

We have called on functions and subroutines in the NAG library for certain standard calculations and you will need to substitute these if you do not have access to the NAG library. We have created simple functions and subroutines which do nothing other than call the NAG library so that there is just one place to make each replacement.

We would appreciate evidence of any errors or difficulties in using the programs -- but note we only claim to provide subroutines and program templates to help you in writing your own programs. More user-friendly software is available commercially in the packages EaSt and PEST.

We are extending this collection of programs as time permits. We hope, eventually, to include routines for implementing all the key methods reported in the book. If you have a particular request please send email to c.jennison@bath.ac.uk -- but understand that we cannot make any firm promises.

To retrieve subroutines or a main program

Click on the highlighted name. A listing will appear in the window.
Click on File (top left hand corner) and select the Save option to save as a FORTRAN file -- in text format. The program should be ready to compile and use.

Programs for normal response variables

Subroutines a collection of subroutines (all together in a single file)
- GST1 calculates properties of a specified group sequential one-sided test.
- GST2 calculates properties of a specified group sequential two-sided test.
- ERSP2 creates boundaries for a two-sided error spending test.
- NORCDF returns the cumulative distribution of a standard normal distribution.
- INVNOR returns a quantile of the standard normal distribution.
Main programs
- MAIN1 creates one-sided Pampallona & Tsiatis tests and calls the subroutine GST1 to evaluate the tests' properties.
- MAIN2 creates two-sided boundaries of Pocock, O'Brien & Fleming and, Wang & Tsiatis tests and calls the subroutine GST2 to evaluate the tests' properties.
- MAINES2 creates two-sided boundaries using an error spending function and calls the subroutine GST2 to evaluate the tests' properties.

Programs for binomial response variables

Subroutines a collection of subroutines (all together in a single file)
- BGST calculates exit probabilities through upper and lower boundaries for a group sequential test with Bernoulli responses.
- BPCI calculates probabilities of a group sequential test for Bernoulli responses terminating in a more extreme state than a particular observed outcome.
- BPRCAL calculates the distribution and cumulative distribution function of a Binomial(n,p) random variable.
- FINDCI calculates a confidence interval for a Bernoulli success probability on termination of a group sequential test.
Main programs
- BMAIN1 produces a table of confidence intervals for outcomes of a 7 stage test taken from MIL STD 105D, plan Q 1.0.
- BMAIN2 produces a table of confidence intervals for outcomes of a 3 stage test taken from Fleming (Biometrics, 1982).
- BMAIN3 calculates error rates and expected sample sizes under two values of the success probability for a 3 stage test taken from Fleming (Biometrics, 1982).
- BMAIN4 produces a table of confidence intervals for outcomes of a 2 stage test with early stopping at stage 1 only for a low number of successes.