 Usage:  locpoldis(x, xg, y, h, logi {, xd})  
 
 Input:

  x                      n x d matrix nonlinear inputs 
                         
  xg                     m x d matrix grid points at which the estimation will be done 
                         
  y                      n x q matrix dependent variable 
                         
  h                      d x 1 matrix vector of bandwidthes 
                         
  logi                   logical scalar 0 if linear, 1 if quadratic polynomials are wished 
                         
  xd                     n x dis matrix optional, inputs for the linear part 
                         
 Output:

  mh                     m x dd x q matrix regression of the nonlinear part and its v'th derivatives 
                         divided by v at xg in the following order: 
                         regression function, 
                         first derivatives respectively x1, x2, x3, ..., 
                         second derivative*0.5 respectively x1, x2, x3, ... 
                         
                         
--------------------------------------------------------------
(C) MD*TECH Method and Data Technologies, 17.8.2000
