 Usage:  {A,b,V,ctrl}=ICerzsep(e,S1,S2,A0,b0,N,eps,itmax,expl,fact0,aus)  

 

 Input:



  e                      numeric; efficiency loss to attain 

                         

  S1                     p x p; Covariance of the first (clipped) component 

                         

  S2                     p x p; Covariance of the second (unclipped) component 

                         

  A0                     p x p; starting value for A; if 0 is entered I^{-1} is taken 

                         else if p>1 and dim(A0)==1 FI^{-1}*A0 is taken 

                         and if p==1 and A0<0 , -A0/FI is taken 

                         

  b0                     numeric; starting value for clipping height; if 0 is entered 

                         4*max(vec(abs("A0"))) [i.e. perhaps modified] 

                         

  N                      integer; MC-sample size / integration grid-points 

                         

  eps                    numeric; exactitude 

                         

  itmax                  integer; maximal number of ICerzseptions 

                         

  expl                   numeric; threshold for the changes in abs. value of A: 

                         beyond this value convergence is uncertain ~ 4 

                         

  fact0                  numeric (>1)!; factor determining how fast we descend from b to b 

                         

  aus                    integer; 0: no output during execution, 1: some output, 2: more output, 3: a lot of output 

                         

 Output:



  A                      p x p; Lagrange-Multiplyer solving (1) 

                         

  b                      1; clipping height 

                         

  V                      p x p; corresponding Covariance 

                         

  ctrl                   integer; tells if convergence "happened" 

                         

--------------------------------------------------------------

(C) MD*TECH Method and Data Technologies, 21.9.2000

