Usage: 
mh = regxestp(x {,h {,K} {,v} })

Input: 
 x  n x (k+1) matrix, the data. In the first k columns
the independent variables is contained and in the
last column the dependent one.

 h  optional scalar, k x 1 vector or 1 x k vector, bandwidth. If not
given, 20% of the range of x[,1:k] is used as default.

 K  optional string, kernel function on [1,1] or Gaussian
kernel "gau". If not given, the Quartic kernel
"qua" is used as default.

 v  optional m x k matrix, values of the independent variable in
which to compute the regression.
If not given and k < 4, a grid of length 100 (k = 1),
length 30 (k = 2) or length 8 (k = 3) is used.
If k >= 4 then v is set to the (sorted) x.

Output: 
 mh  n x (k+1) or m x (k+1) matrix, the first k columns
contain the grid or the sorted x[,1:k], the
last column contains the regression estimate
on the values of the first k columns. 