nlsr (version 2023.8.31)

numericDerivR: numericDerivR: numerically evaluates the gradient of an expression. All in R

Description

This version is all in R to replace the C version in package stats

Usage

numericDerivR(
  expr,
  theta,
  rho = parent.frame(),
  dir = 1,
  eps = .Machine$double.eps^(1/if (central) 3 else 2),
  central = FALSE
)

Value

The value of eval(expr, envir = rho) plus a matrix attribute "gradient". The columns of this matrix are the derivatives of the value with respect to the variables listed in theta.

Arguments

expr

expression or call to be differentiated. Should evaluate to a numeric vector.

theta

character vector of names of numeric variables used in expr.

rho

environment containing all the variables needed to evaluate expr.

dir

numeric vector of directions, typically with values in -1, 1 to use for the finite differences; will be recycled to the length of theta.

eps

a positive number, to be used as unit step size hh for the approximate numerical derivative (f(x+h)-f(x))/h (f(x+h)-f(x))/h or the central version, see central.

central

logical indicating if central divided differences should be computed, i.e., (f(x+h) - f(x-h)) / 2h (f(x+h)-f(x-h))/2h. These are typically more accurate but need more evaluations of f()f().

Examples

ex <- expression(a/(1+b*exp(-tt*c)) - weed)
weed <- c(5.308, 7.24, 9.638, 12.866, 17.069, 23.192, 31.443,
              38.558, 50.156, 62.948, 75.995, 91.972)
tt <- 1:12
a <- 200; b <- 50; c <- 0.3
dhobb <- numericDerivR(ex, theta=c("a", "b", "c"))
print(dhobb)
# exf <- ~ a/(1+b*exp(-tt*c)) - weed
# Note that a formula doesn't work
# dh1 <-  try(numericDerivR(exf, theta=c("a", "b", "c")))