University of Texas M. D. Anderson Cancer Center
Department of Biomathematics
AVAILABLE SOFTWARE
December 31, 1990
Barry W. Brown
Department of Biomathematics, Box 237
University of Texas M. D. Anderson Cancer Center
1515 Holcombe Boulevard
Houston, TX 77030
bwb@mdaali.cancer.utexas.edu
I. WHAT IS AVAILABLE?
KSB1CI
KSB1CI (k-stage binomial one group confindence intervals) finds
confidence intervals for single- or multiple-stage trials involving a
single group which have a binomial outcome. The confidence intervals
are calculated by a minor modification of the Clopper and Pearson
method to k-stage use. The lower value of the confidence limit is
that probability that causes the observed number of events or more to
have probability 1-c/2 where c is the confidence level; the upper
limit is symmetrically defined.
Reference
Clopper, C.J. and Pearson, E.S. (1934) "The use of confidence or
fiducial limits illustrated in the case of the binomial," Biometrika,
26, 404.
KSBIN1
KSBIN1 performs computations needed to plan multiple-stage experiments
involving dichotomous outcomes for a single group; the outcomes are
distributed binomially. In k-stage experiments, the results are
examined at each of k predetermined times; at each examination, the
experiment can be terminated because of either early proof of the
result or because obtaining such proof is very unlikely if the
experiment is continued. The assumption of a binomial distribution of
outcomes implies that the outcomes are independent of each other at
that the probability of the event of interest is the same for each
outcome. KSBIN1 performs its calculations only for a single group;
i.e., the case in which a comparison of the probability of event with
a fixed value.
RANLIST
RANLIST produces randomization lists and offers a primitive form of
interactive randomization. RANLIST features both restricted and
unrestricted randomization. When restricted randomization is used,
fixed or random balance points may be chosen.
STATTAB
STATTAB calculates the cumulative distribution funcions, inverses, and
values of parameters for many common statistical distributions. It
can also produce tables of any one calculated iten as a function of a
specified list of values of another item.
STPLAN
STPLAN performs power, sample size, and related calculations needed to
plan studies. Parameters of a test include those necessary to
describe the distributions under the null and alternative hypotheses,
sample sizes, significance levels, and power. Given values for all
but one parameter, STPLAN will calculate the value of that one
parameter.
WINDOWS
WINDOWS provides a variety of window/kernel/moving-average estimators.
Both fixed width and fixed number of nearest neighbor windowing is
provided. Weighting of points may be done either with boxcar or
biquartic weights. Estimates available include weighted mean, a
linear or quadratic fit, maximum or minimum value, a specified
percentile, the standard deviation, and the derivative of y with
respect to x. Cross validation is also included.
II. WHAT IS THE POLICY ON USE OF THIS CODE?
We want to make these programs as widely available as possible.
Hence, executable code derived from the source provided, the source
itself, and the accompanying documentation may be freely copied and
distributed provided that the copyright accompanies such copies. We
encourage the distribution of this code on bulletin boards and as
shareware.
Modifications and improvements to the code can similarly be freely
copied and distributed. Barry Brown would appreciate a copy of any
such improvements for incorporation into later versions.
Source code may be extracted and used for any non-commercial
purpose. Non-commercial means that at most a modest distribution fee
is charged, as shareware. We would appreciate acknowledgment of usch
use of our code.
Commercial use of the source code requires written permission from
the Department of Biomathematics at the above address. Our policy is
generally to grant such permission provided appropriate credit is
given to us in the documentation.
THERE IS NO WARRANTY ON ANY OF OUR CODE.
III. HOW CAN YOU GET THIS CODE?
Perhaps the best way to obtain this code is to copy it from
someone who has it.
The second best method is to obtain it by anonymous ftp from
mdaali.cancer.utexas.edu (129.106.1.6). To do so, ftp to this machine
and log on as user anonymous. Any password will work. The source and
documentation for each package is on a compressed tar file
.tar.Z. (There is one exception. An abbrieviated version of
the STPLAN documentation is packaged with the code. The abbrieviated
version should suffice for statisticians. The full documentation is
on stplan.doc.Z on /pub.) Use the 'type binary' command before
retrieving files. Each package should be restored in a separate
subdirectory as the packages have commonly named files. To restore
the files, the following commands should be used:
uncompress .tar.Z
(This creates an uncompressed file, .tar.)
tar xf .tar
(This recreates the original files.)
(Read file README and proceed from there.)
Directories /pub/msdos and /pub/mac contain MSDOS and MacIntosh
executables also packaged as tar.Z files. These files must be
restored in the same fashion as the source files. Then the executable
files must be downloaded from the Unix machine to the PC or Mac.
Hints for doing this are given in the README files on the two
subdirectories.
Source code and documentation is or shortly will be posted on
statlib where those with access to internet or bitnet can obtain it by
electronic mail. To obtain information on statlib, send electonic
mail whose content is 'send index' to statlib@lib.stat.cmu.edu. The
code described is in subsection general whose index can be obtained by
a message 'send index from general' to the same address.
Since potential researchers with microcomputers may well not have
a FORTRAN compiler, we will place MacIntosh and MSDOS executables and
documentation on CompuServe if they accept it and elsewhere.
(Elsewhere will be described when we figure out where it will be.) On
CompuServe Go IBMPRO for MSDOS and GO MACAPP for MacIntosh versions.
If all else fails, the code can be ordered from us. MacIntosh or
MSDOS executables plus electronic documentation costs $150.00 and is
available on:
MacIntosh
800 KB 3.5" floppies
MSDOS
1.44 MB 3.5" floppies
720 KB 5.25" floppies
360 KB 5.25" floppies
Full source for the programs (the same source available from
statlib) can be ordered for $300.00. Source is available on the same
media as the executables for MacIntosh and MSDOS machines. In
addition, the source is available on 9 track tapes in BACKUP format
for VAX machines and tar format for UNIX machines.