Usage: 
{f,variance}=lplocband(x,y,h,xgrid,OrderDer,p{,RidgeCoef{,Kernel}})

Input: 
 x  n x k matrix, the independent variable.

 y  n x 1 vector, the dependent variable

 h  bandwidth which may be local or global. Possible dimensions:
 m x k matrix (local bandwidths)
 1 x k vector (global bandwidths)
 scalar (global bandwidth)

 xgrid  m x k matrix, grid where you
estimate the dependent variable.

 OrderDer  scalar, order of the derivative you want to estimate. This order
depends on the dimension of the independent variables.
 OrderDer = 0 corresponds always to the function itself
 1 <= OrderDer <= k corresponds to the first derivatives
 k < OrderDer <= 2*k corresponds to the non mixed
second derivatives
 OrderDer > 2*k corresponds to the mixed derivatives

 p  scalar, degree of the polynomials (1<=p<=2). Take care that if you want to
estimate second derivatives, p must be equal to 2.

 RidgeCoef  optional scalar representing the Ridge coefficient if ridge regression
is desired.

 Kernel  optional string defining the kernel function used. The kernel functions
available are Quartic ("Qua"), Epanechnikov ("Epa") and Triangle ("Tri")

Output: 
 f  m x 1 vector representing the estimated OrderDerth derivative of
E(yx) at each point on xgrid. 
 variance  m x 1 vector containing the estimate of the variance of this estimate
divided by sigma2, where sigma2 is the variance of the y's. 