logbin (version 1.0)

B.Iso: Defining Smooths in logbin.smooth Formulae

Description

Function used in the definition of smooth terms within logbin.smooth model formulae. The function does not evaluate a smooth - it exists purely to help set up a model using smooths.

Usage

B(..., knots = NULL, knot.range = 0:5)

Iso(...)

Arguments

...

variable that this smooth is a function of. Note that unlike gam, smooths that are functions of more than one variable are not supported.

knots

unique positions of interior knots of a B-spline basis. Boundary knots are created automatically.

knot.range

if knots is not specified, a vector containing a series of non-negative integers denoting the number of interior knots for which the model will be fit. These are placed at evenly-spaced quantiles of the observed covariate values.

Value

• An object of class "B.smooth" (for B) or "Iso.smooth" (for Iso), which is a list with the following elements:
• term
  name of the term provided in the ... argument.
• termlabel
  label for the term in the model; e.g. for term "x" it will be "B(x)" or "Iso(x)".
• knots
  vector of interior knots (if specified). NA for Iso.
• knot.range
  vector of number of interior knots. NA for Iso.

Details

The function does not evaluate the variable arguments; the output from this function is passed as part of the arguments to logbin.smooth.design, which constructs the actual basis functions. B is used to specify an order-3 B-spline basis (which can be restricted to be monotonically non-decreasing via the mono argument in logbin.smooth). If length(knot.range) > 1, models with each of the specified number of interior knots will be fit, and the model with the best (smallest) aic.c will be returned. Iso is used to specify an isotonic basis, designed such that the resulting function has non-negative increments at each observed covariate value. When Iso is used, the resulting function will always be monotonically non-decreasing, regardless of the value of mono.

See Also

logbin.smooth, logbin.smooth.design s performs a similar function in the mgcv package.

Examples

## See example(logbin.smooth) for an example of specifying smooths in model
##  formulae.