 Usage:  b = lts(x, y{, h, all, mult})   
 
 Input:

  x                      n x p design matrix of explanatory variables. 
                         
  y                      n x 1 vector, dependent variable. 
                         
  h                      optional, trimming constant; default value is 
                         [n/2]+[(p+1)/2], where n is the number of observations 
                         and p is the number of parameters. h should belong to 
                         {[(n+1)/2],...,n} and must be bigger than p. 
                         
  all                    optional, logical flag for the exact computation (nonzero = TRUE), 
                         default value is 0 (FALSE). If ci = 0, an approximation 
                         of the estimator is computed. If ci != 0, the estimate 
                         is computed precisely - this can be rather time demanding 
                         and applicable only for small sample sizes. The number of 
                         iterations corresponds in this case to n over h. 
                         
  mult                   optional, affects the maximal number of iterations, 
                         after which the algorithm is stoped (if convergence 
                         was not achieved earlier), default value equals to 1. 
                         The maximal number of iterations is currently 
                         '(600 * h) * times', so this variable allow 
                         to reduce or extend this number. 
                         
                         
 Output:

  b                      p x 1 vector of estimated coefficients. 
                         
--------------------------------------------------------------
(C) MD*TECH Method and Data Technologies, 17.8.2000
