from_ped_to_tensor: Transform PED data to an array / tensor

In pkopper/deeppam: Machine Learning for Time-to-Event Data with Semi-Structured Deep Piecewise Additive Models

Description

This function transforms PED data (from pammtools) to an array or tensor for further usage in tensorflow. The array is constructed from the PED so that the first dimension has same length as there are unique IDs in the PED. The second dimension reflects all distinct intervals per ID and the third dimension features all model features effectively used by the PAM(M). The third axis features the offset ([, , 1]) all linear covariates and all spline covariates in a already transformed manner using mgcv::gam for the transformation. The transformation is based on the formula argument.

Usage

from_ped_to_tensor(
  ped,
  formula = ped_status ~ s(tend, k = 9L, bs = "cr"),
  training = TRUE,
  training_data = NULL,
  id = "id"
)

Arguments

ped	a PED (see pammtools) data frame.
formula	a formula for a PAM(M) (see pammtools/mgcv)
training	a boolean indicating if a training or test set is to be constructed.
training_data	only relevant if training == TRUE. The training PED that has been used to fit the model is to be tested.
id	a string indicating the name of the ID variable.

Value

A list of class ped_tensor and length 2. [[1]] is the associated input (X) and [[2]] the associated target (y).

Author(s)

Philipp Kopper

pkopper/deeppam documentation built on Jan. 19, 2021, 12:39 a.m.