genGradient: Compute Finite Difference Gradient and Hessians.
In SpatioTemporal: Spatio-Temporal Model Estimation

Description Usage Arguments Value Author(s) See Also Examples

Computes finite difference gradient and/or hessian. genGradient function does forward, backward or central differences, the genHessian function uses only central differences.

1
2
3

genGradient(x, func, h = 0.001, diff.type = 0)

genHessian(x, func, h = 0.001)

`x`	Point at which to compute the gradient or hessian.
`func`	function that takes only `x` as an input argument. Use `function(x){my.func(x,other.input)}` to create a temporary function, see the example.
`h`	Step length for the finite difference.
`diff.type`	Type of finite difference, `diff.type>0` gives forward differences, `diff.type=0` gives central differences, and `diff.type<0` gives backward differences.

gradient vector or Hessian matrix.

Johan Lindstrom

Other numerical derivatives: loglikeSTGrad

#create a two variable function
f.test <- function(x){sin(x[1])*cos(x[2])}

#compute the gradient using forward difference
genGradient(c(.5,.5), f.test, diff.type=1)
#and central difference
genGradient(c(.5,.5), f.test, diff.type=0)
#compared to the true value
c(cos(.5)*cos(.5),-sin(.5)*sin(.5))

#Compute the Hessian
genHessian(c(.5,.5), f.test, h=1e-4)
#and compare to the true value
matrix(c(-sin(.5)*cos(.5),-cos(.5)*sin(.5),
         -cos(.5)*sin(.5),-sin(.5)*cos(.5)),2,2)