BoxMeyer.exe (Windows):

This uses a Bayesian procedure to find the best 2^1, 2^2, and 2^3
  submodels of an experimental design under various criteria.
  See the file  BayesRegrBoxMeyer.pdf  for references and
  a description of the theory.

Usage of BoxMeyer.exe:

For example from a command line:

    BoxMeyer Myfile.txt

  Reads an input file in a format like PBExamp1.txt or FF84Ex2.txt
    in the same directory as this file
    See PBExamp1_doc.txt for a description of the input-file format
  Writes an output file  Myfile.txt.out.txt  or Myfile.out.txt

If you ``Drag-and-Drop'' Myfile.txt on top of `BoxMeyer.exe' in Windows
  Explorer, the effect is the same as entering `BoxMeyer Myfile.txt' on a
  command line. The program will then write a file  `Myfile.txt.out.txt'
  in the same directory as BoxMeyer.exe and Myfile.txt.

(See below for more information about using BoxMeyer.exe.)

See PBExamp1.out.txt and FF84Ex2.out.txt for example output files.
  If successful, BoxMeyer writes to the output file:

    (i) Various descriptive information about this run,
    (ii) A list of the top (up to) 5 models for each model size
      (sizes=1,2,3) as sorted by the F-statistic, along with other
      information,
    (iii) A list of the top (up to) 15 models for all model sizes
      sorted by the F-statistic,
    (iv) A list of the top (up to) 15 models for all model sizes
      sorted by the the Bayesian posterior probability,
    (v) Posterior probabilities for factors in the model that the
      factor is active, with models weighted by the F-statistic
    (vi) Posterior probabilities for factors in the model that the
      factor is active, with models weighted by Bayesian posterior
      probability.

For fractional factorial models, weighting and sorting by the F-statistic
  seems to work better than the Bayesian posterior probability. For
  Plackett-Burman designs, the Bayesian procedure seems to work better.

Submodels whose design matrix is not invertible are ignored. (That is,
  the program will not crash). For example, this will happen if the
  program tries to analyze a 2^3 submodel for columns that are equivalent
  to A B AB, since then the two-way interaction of the first two columns
  is identical with the third column. Non-invertible models are sorted
  last in any list of models.

More about the program:

BoxMeyer.exe reads parameter settings from the second non-blank
  non-comment line of Myfile.txt (see PBExamp1_doc.txt). Parameters
  can also be set from the command line. Enter `BoxMeyer' by itself in a
  console (or DOS-Prompt) window to see the possible command-line
  options. For example,

    BoxMeyer -p0.35 -g7 Myfile.txt

  runs the program with Pi=0.35 (default value: 0.25) and Gamma=7
  (default value: 2.5). Command-line arguments take precedence over
  settings in Myfile.txt. As another example,

    BoxMeyer -l Myfile.txt

  runs the program with NV=Ncols (see PBExamp1_doc.txt). Command-line
  arguments including Myfile.txt can be entered in any order. If you
  enter a second file name on the command line (that is, something after
  Myfile.txt that doesn't begin with -), then that becomes the name of
  the output file.


For legal purposes, no warranty or guarantees are made about how the
  program will work or if it will work.

Stanley Sawyer
Department of Mathematics
Washington University
St.Louis, MO 63130
email:  sawyer@math.wustl.edu
Voice:  314-935-6703
FAX:    314-935-6839  (departmental FAX)
Web:    http://www.math.wustl.edu/~sawyer