survex (version 1.2.0)

model_profile_2d: Dataset Level 2-Dimensional Variable Profile for Survival Models

Description

This function calculates explanations on a dataset level that help explore model response as a function of selected pairs of variables. The explanations are calculated as an extension of Partial Dependence Profiles or Accumulated Local Effects with the inclusion of the time dimension.

Usage

model_profile_2d(
  explainer,
  variables = NULL,
  N = 100,
  categorical_variables = NULL,
  grid_points = 25,
  center = FALSE,
  variable_splits_type = "uniform",
  type = "partial",
  output_type = "survival"
)
# S3 method for surv_explainer
model_profile_2d(
  explainer,
  variables = NULL,
  N = 100,
  categorical_variables = NULL,
  grid_points = 25,
  center = FALSE,
  variable_splits_type = "uniform",
  type = "partial",
  output_type = "survival"
)

Value

An object of class model_profile_2d_survival. It is a list with the element result containing the results of the calculation.

Arguments

explainer

an explainer object - model preprocessed by the explain() function

variables

list of character vectors of length 2, names of pairs of variables to be explained

N

number of observations used for the calculation of aggregated profiles. By default 100. If NULL all observations are used.

categorical_variables

character, a vector of names of additional variables which should be treated as categorical (factors are automatically treated as categorical variables). If it contains variable names not present in the variables argument, they will be added at the end.

grid_points

maximum number of points for profile calculations. Note that the final number of points may be lower than grid_points. Will be passed to internal function. By default 25.

center

logical, should profiles be centered around the average prediction

variable_splits_type

character, decides how variable grids should be calculated. Use "quantiles" for quantiles or "uniform" (default) to get uniform grid of points. Used only if type = "partial".

type

the type of variable profile, "partial" for Partial Dependence or "accumulated" for Accumulated Local Effects

output_type

either "survival", "chf" or "risk" the type of survival model output that should be considered for explanations. If "survival" the explanations are based on the survival function. If "chf" the explanations are based on the cumulative hazard function. Otherwise the scalar risk predictions are used by the DALEX::predict_profile function.

Examples

# \donttest{
library(survival)
library(survex)

cph <- coxph(Surv(time, status) ~ ., data = veteran, model = TRUE, x = TRUE, y = TRUE)
cph_exp <- explain(cph)

cph_model_profile_2d <- model_profile_2d(cph_exp,
    variables = list(c("age", "celltype"))
)
head(cph_model_profile_2d$result)
plot(cph_model_profile_2d)

cph_model_profile_2d_ale <- model_profile_2d(cph_exp,
    variables = list(c("age", "karno")),
    type = "accumulated"
)
head(cph_model_profile_2d_ale$result)
plot(cph_model_profile_2d_ale)
# }