Maximum likelihood estimation for the Generalized Box-Cox transformation.

powFit(x, alpha = NA, interval = c(-5, 5), ...)

Arguments

x

Vector of random samples from target density.

alpha

Optional value of the offset parameter. alpha = FALSE sets alpha = 1 - min(x), thereby guaranteeing that z = x + alpha >= 1. This or any scalar value of alpha finds the conditional MLE as a function of lambda only. alpha = NA finds the joint MLE over (lambda,alpha).

interval

Range of lambda values for one dimensional optimization.

...

Additional arguments to pass to optimize or optim, for 1- or 2-parameter optimization.

Value

Vector of length two containing the fitted and/or known values of (lambda, alpha).

Details

The likelihood for optimization is

L(lambda, alpha | x) = prod(dnorm(z(x | lambda, alpha)) *  |dz(x | lambda, alpha) / dx|)

, where z(x | lambda, alpha) is the Box-Cox transformation.