The t_α transformation as a link function for binary GLMs.

Share:

Description

A parametric link function, i.e. a family of link functions intended for binary data.

Usage

1
2
talpha(alpha, verbose = FALSE,
  splineinv = TRUE, eps = 2 * .Machine$double.eps, maxit = 100)

Arguments

alpha

numeric, must be larger than 0 and smaller than 2.

verbose

logical, warn if truncation occurs when link function or inverse are used.

splineinv

logical, use spline interpolation for calculation of inverse link?

eps

if splineinv is FALSE, a Newton-Raphson algorithm is run to calculate the inverse. The argument eps determines when to terminate this algorithm. Ignored if splineinv is TRUE.

maxit

maximum number of iterations for Newton-Raphson. Ignored if splineinv is TRUE.

Value

An object of class "link-glm", see family and family. Intended for use with glm.

Author(s)

Philipp Doebler <philipp.doebler@googlemail.com>

Examples

1
2
3
canonical <- binomial(link = talpha(1)) # logit-link
talpha_fam <- function(alpha)binomial(link = talpha(alpha)) # talpha family
## A call to glm might look like this: glm(formula, family = talpha_fam(1.5))