RaphaelS1

name: Documentation

on:
  push:
    branches:
      - main
      - master
    tags: '*'
  pull_request:

RaphaelS1

#' Get SHAPs from a trained xgboost model
#' @param model trained xgboost model
#' @param data new data for calculating shaps
get_shaps <- function(model, data) {
  p = predict(model, data,
              predcontrib = T, approxcontrib = F)
  p = p[, -ncol(p)]

  shaps = reshape2::melt(p)
  vars = reshape2::melt(data)

RaphaelS1

`%=%` <- function(l, r) {
    l <- trimws(strsplit(l, ",", TRUE)[[1]])
    if (all(l == "USE.NAMES") || all(l == "*") || all(l == "?")) {
      stopifnot(length(names(r)) > 0)
      l <- names(r)
    } else {
      stopifnot(identical(length(l), length(r)))
      which <- l == "?"
      if (any(which)) {
        stopifnot(length(names(r)) > 0)

RaphaelS1

raincloudplot <- function(data, y, fill, x = fill, cols = NULL,
                          boxplots = c("below", "in", "none")) {
  
  require(dplyr)
  require(ggplot2)
  source("https://raw.githubusercontent.com/datavizpyr/data/master/half_flat_violinplot.R")
  
  boxplots <- match.arg(boxplots)

  p <- ggplot(data, aes_string(y = y, x = x, fill = fill)) +

RaphaelS1

## load ggplot2 for autoplots
library(ggplot2)

## critical difference diagrams for IGS
autoplot(bma, meas = "graf", type = "cd", ratio = 1/3, p.value = 0.1)

RaphaelS1

library(mlr3benchmark)

## create mlr3benchmark object
bma <- as.BenchmarkAggr(bm)

## run global Friedman test
bma$friedman_test()

RaphaelS1

## select holdout as the resampling strategy
resampling <- rsmp("cv", folds = 3)

## add KM and CPH
learners <- c(learners, lrns(c("surv.kaplan", "surv.coxph")))
design <- benchmark_grid(tasks, learners, resampling)
bm <- benchmark(design)
               
## Aggreggate with Harrell's C and Integrated Graf Score
msrs <- msrs(c("surv.cindex", "surv.graf"))

RaphaelS1

library(mlr3pipelines)

create_pipeops <- function(learner) {
 po("encode") %>>% po("scale") %>>% po("learner", learner)
}

## apply our function
learners <- lapply(learners, create_pipeops)

RaphaelS1

## learners are stored in mlr3extralearners
library(mlr3extralearners)

## load learners
learners <- lrns(
 paste0("surv.", c("coxtime", "deephit", "deepsurv", "loghaz", "pchazard")),
 frac = 0.3, early_stopping = TRUE, epochs = 10, optimizer = "adam"
)
 
# apply our function

RaphaelS1

library(mlr3tuning)

create_autotuner <- function(learner) {
 AutoTuner$new(
   learner = learner,
   search_space = search_space,
   resampling = rsmp("holdout"),
   measure = msr("surv.cindex"),
   terminator = trm("evals", n_evals = 2),
   tuner = tnr("random_search")