New SEQBIN - A Bayesian/Frequentist design for sequential trials withbinomial outcome

Discussion in 'Scientific Statistics Math' started by Barry W Brown, Dec 1, 2011.

  1. ANNOUNCE: SEQBIN-1,5 (Sequential Binomial Trial)

    DESCRIPTION

    This program implements a combined Bayesian and frequentist strategy
    for designing sequential and group sequential trials that test a
    one-group hypothesis on a probability. (For example, the cure rate in
    a Phase II trial.) The user specifies a Bayestian prior distribution
    on this probability and a stopping probability, S. The trial stops if
    the posterior probability that the actual value is greater than the
    null value (say the known cure rate without treatment) is at least S.
    If the trial stops for this reason, the treatment is judged to be worthy
    of further testing or else of general use.

    HOW TO GET IT

    The program is free (no charge, Fortran 95 source included) and is
    available from the following URL. The distribution includes source,
    documentation, a Win32 executable and a MAC executable.

    http://biostatistics.mdanderson.org/SoftwareDownload

    Note: There is a lot of stuff there. Use the search box for 'seqbin'
    or click on the column heading, 'Product Name' to sort the list of
    names.

    WHAT'S NEW?

    A Mac executable for the current (OSX) operating system has been
    added. The documentation has been improved.

    EXAMPLE:

    The output is too voluminous to include here. An actual use is the
    following. A trial was designed to test the response of spinal cord
    cancer to radiation. Paralysis was a possible and extremely
    undersired effect of the treatment. A sequential design was adopted
    that would terminate the trial if there was evidence at the 0.05 level
    that the incidence of paralysis was greater than 2%. This stopping
    condition was imposed on a different design for effectiveness.
     
    Barry W Brown, Dec 1, 2011
    #1
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.