 Usage:  {fh,c} = interact(t,y,h,g,loc,incl{,tg})  
 
 Input:

  t                      n x p matrix , the observed explanatory variable 
                         
  y                      n x 1 vector , the observed 
                         response variables 
                         
  h                      p x 1 or 1 x 1 matrix , chosen bandwidth for 
                         the directions of interest in the estimation step 
                         
  g                      p x 1 or 1 x 1 matrix , chosen bandwidth for 
                         the directions not of interest in the estimation step 
                         
  loc                    scalar in {0,1,2}, the degree of the local polynomial 
                         smoother 
                         
  incl                   pp x 2 matrix, in the rows have to be all pairs 
                         of indices for interactions which shall be included 
                         in the model. 
                         
  tg                     ng x p matrix , if grid is wished, tg is the grid on 
                         which we will estimate. Attention ! If you estimate on 
                         a grid, the estimates of the interaction functions 
                         are estimated up to a constant shift. Further, if you 
                         try to extrapolate, the estimate will be zero minus 
                         a centering constant. 
                         
 Output:

  mod.fh                 n x (p+pp) matrix, estimates for the functions 
                         
  mod.c                  scalar, the overall constant 
                         
--------------------------------------------------------------
(C) MD*TECH Method and Data Technologies, 17.8.2000
