pracma (version 1.1.6)

fderiv: Numerical Differentiation

Description

Numerical function differentiation for orders n=1..4 using finite difference approximations.

Usage

fderiv(f, x, n = 1, h = 0,
       method="central", ...)

Arguments

f

function to be differentiated.

x

point(s) where differentiation will take place.

n

order of derivative, can only be between 1 and 4; for n=0 function values will be returned.

h

step size: if h=0 step size will be set automatically.

method

one of ``central'', ``forward'', or ``backward''.

...

more variables to be passed to function f.

Value

• Vector of the same length as x.

Details

Derivatives are computed applying central difference formulas that stem from the Taylor series approximation. These formulas have a convergence rate of $O(h^2)$.

Use the `forward' (right side) or `backward' (left side) method if the function can only be computed or is only defined on one side. Otherwise, always use the central difference formulas.

Optimal step sizes depend on the accuracy the function can be computed with. Assuming internal functions with an accuracy 2.2e-16, appropriate step sizes might be 5e-6, 1e-4, 5e-4, 2.5e-3 for n=1,...,4 and precisions of about 10^-10, 10^-8, 5*10^-7, 5*10^-6 (at best).

References

Kiusalaas, J. (2005). Numerical Methods in Engineering with Matlab. Cambridge University Press.

See Also

numderiv, taylor

Examples

f <- sin
xs <- seq(-pi, pi, length.out = 100)
ys <- f(xs)
y1 <- fderiv(f, xs, n = 1, method = "backward")
y2 <- fderiv(f, xs, n = 2, method = "backward")
y3 <- fderiv(f, xs, n = 3, method = "backward")
y4 <- fderiv(f, xs, n = 4, method = "backward")
plot(xs, ys, type = "l", col = "gray", lwd = 2,
     xlab = "", ylab = "", main = "Sinus and its Derivatives")
lines(xs, y1, col=1, lty=2)
lines(xs, y2, col=2, lty=3)
lines(xs, y3, col=3, lty=4)
lines(xs, y4, col=4, lty=5)
grid()