terse.txt

                        A Terse Summary of
                            multout
                       A Program for Robust,
                      Multivariate Analysis
                      
THIS SOFTWARE IS PROVIDED "AS IS." THERE IS NO WARRANTY EXPRESS OR IMPLIED 
OF ANY KIND.  MERCHANTABILITY AND FITNESS FOR ANY PARTICULAR USE ARE
EXPRESSLY NOT WARRANTED.

The program attempts to find a robust estimate of shape, location and
Mahalanobis distances in high dimension. The dimension and number of
data points are limited only by computer memory. See the paper
"Identification of Outliers in Multivariate Data" by Rocke and
Woodruff (JASA, September 1996) for detailed explanations of the
algorithms.

To run the program from the command line use:
  multout infile outfile iterations [parmsfile]
where infile contains: p n
                       data record one (p data elements)
                       data record two
                             . . . 
  and iterations is an integer controlling the time spent.
  For light contamination use 1 iteration; for heavy, try n*n

Example: 
  multout suspect.dat suspect.out 10000
  
Example data file for 8 data points in dimension 3 where only
the first two data points are shown:
  3 8
  2.112 1.232 4.9871
  8.665 5.182 3.9982
     
The argument parmsfile is optional and is intended for use only by 
programmers at the present time.

The outfile argument gives the name of a file that will contain
the program output at the end of execution. 

Notes:
  1. The simulations for outlier detection take a long time. If you
     are going to do multiple sets of analysis with simlar data sets,
     and if you are a programmer, you can speed this up a great deal 
     by doing simulations in advance and creating a table
  2. The program cannot deal with singular data very well. A simple and 
     common form of singular data is a large number of data points that are 
     exactly the same. If the program complains about singularities, you may 
     need to consider slight perturbations of your data.  
  3. Send questions and comments to dlwoodruff@ucdavis.edu (916/752-0515)