trafo_control: Options for transformation models

Description

Options for transformation models

Usage

trafo_control(
  order_bsp = 10L,
  support = function(y) range(y),
  y_basis_fun = NULL,
  y_basis_fun_lower = NULL,
  y_basis_fun_prime = NULL,
  penalize_bsp = 0,
  order_bsp_penalty = 2,
  tf_bsps = FALSE,
  response_type = c("continuous", "ordered", "survival", "count"),
  atm_toplayer = function(x) {
     layer_dense(x, units = 1L, name = "atm_toplayer",
    use_bias = FALSE)
 },
  basis = c("bernstein", "ordered", "shiftscale")
)

Value

Returns a named list with all options, basis functions, support, and penalties.

Arguments

order_bsp: The order of Bernstein polynomials in case y_basis_fun is a Bernstein polynomial defined by eval_bsp or (one less than) the number of classes of an ordinal outcome.
support: A function returning a vector with two elements, namely the support for the basis of y.
y_basis_fun: Function; basis function for Y
y_basis_fun_lower: Function; basis function for lower bound of interval censored response
y_basis_fun_prime: Function; basis function derivative
penalize_bsp: Scalar value > 0; controls amount of penalization of Bernstein polynomials.
order_bsp_penalty: Integer; order of Bernstein polynomial penalty. 0 results in a penalty based on integrated squared second order derivatives, values >= 1 in difference penalties.
tf_bsps: Logical; whether to use a TensorFlow implementation of the Bernstein polynomial functions.
response_type: Character; type of response can be continuous, ordered, survival, or count.
atm_toplayer: Function; a function specifying the layer on top of ATM lags.
basis: Character or function; implemented options are "bernstein" (a Bernstein polynomial basis), "ordered" (for ordinal responses), or "shiftscale" for (log-) linear bases