## ----include = FALSE----------------------------------------------------------
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

## ----pipeline-diagram, echo = FALSE, out.width = "100%", fig.cap="Diagram of the landmaRk package pipeline"----
knitr::include_graphics(system.file("diagram.svg", package = "landmaRk"))

## ----setup--------------------------------------------------------------------
set.seed(123)
library(landmaRk)
library(lcmm)
library(tidyverse)
library(prodlim)

## -----------------------------------------------------------------------------
library(JMbayes2)
data("aids")
aids$patient <- as.numeric(aids$patient)
str(aids)

## -----------------------------------------------------------------------------
# DF with Static covariates
aids_dfs <- split_wide_df(
  aids,
  ids = "patient",
  times = "obstime",
  static = c("Time", "death", "drug", "gender", "prevOI"),
  dynamic = c("CD4"),
  measurement_name = "value"
)
static <- aids_dfs$df_static
head(static)

## -----------------------------------------------------------------------------
# DF with Dynamic covariates
dynamic <- aids_dfs$df_dynamic
head(dynamic[["CD4"]])

## -----------------------------------------------------------------------------
landmarking_object <- LandmarkAnalysis(
  data_static = static,
  data_dynamic = dynamic,
  event_indicator = "death",
  ids = "patient",
  event_time = "Time",
  times = "obstime",
  measurements = "value"
)

## -----------------------------------------------------------------------------
landmarking_object <- landmarking_object |>
  compute_risk_sets(landmarks = c(6, 8))

landmarking_object

## -----------------------------------------------------------------------------
landmarking_object <- landmarking_object |>
  fit_survival(
    formula = Surv(event_time, event_status) ~ drug,
    landmarks = c(6, 8),
    horizons = 12 + c(6, 8),
    method = "coxph",
    dynamic_covariates = c()
  ) |>
  predict_survival(
    landmarks = c(6, 8),
    horizons = 12 + c(6, 8)
  )

## -----------------------------------------------------------------------------
summary(landmarking_object, type = "survival", landmark = 6, horizon = 18)

## -----------------------------------------------------------------------------
performance_metrics(
  landmarking_object,
  landmarks = c(6, 8),
  horizons = c(18, 20),
  auc_t = TRUE, c_index = FALSE,
  h_times = c(3, 6, 12)
)

## -----------------------------------------------------------------------------
landmarking_object <- LandmarkAnalysis(
  data_static = static,
  data_dynamic = dynamic,
  event_indicator = "death",
  ids = "patient",
  event_time = "Time",
  times = "obstime",
  measurements = "value"
)
landmarking_object <- landmarking_object |>
  compute_risk_sets(landmarks = c(6, 8))

## -----------------------------------------------------------------------------
landmarking_object <- landmarking_object |>
  fit_longitudinal(
    landmarks = c(6, 8),
    method = "lme4",
    formula = value ~ prevOI + obstime + (obstime | patient),
    dynamic_covariates = c("CD4")
  ) |>
  predict_longitudinal(
    landmarks = c(6, 8),
    method = "lme4",
    dynamic_covariates = c("CD4")
  ) |>
  fit_survival(
    formula = Surv(event_time, event_status) ~ drug,
    landmarks = c(6, 8),
    horizons = 12 + c(6, 8),
    method = "coxph",
    dynamic_covariates = c("CD4")
  ) |>
  predict_survival(
    landmarks = c(6, 8),
    horizons = 12 + c(6, 8)
  )

## -----------------------------------------------------------------------------
summary(landmarking_object,
        type = "longitudinal",
        landmark = 6,
        dynamic_covariate = "CD4")

## -----------------------------------------------------------------------------
summary(landmarking_object, type = "survival", landmark = 6, horizon = 18)

## -----------------------------------------------------------------------------
performance_metrics(
  landmarking_object,
  landmarks = c(6, 8),
  horizons = c(18, 20),
  auc_t = TRUE, c_index = FALSE,
  h_times = c(3, 6, 12)
)

## -----------------------------------------------------------------------------
landmarking_object <- LandmarkAnalysis(
  data_static = static,
  data_dynamic = dynamic,
  event_indicator = "death",
  ids = "patient",
  event_time = "Time",
  times = "obstime",
  measurements = "value"
)

## -----------------------------------------------------------------------------
landmarking_object <- landmarking_object |>
  compute_risk_sets(landmarks = c(6, 8)) |>
  fit_longitudinal(
    landmarks = c(6, 8),
    method = "lcmm",
    formula = value ~ obstime + prevOI,
    mixture = ~ obstime + prevOI,
    random = ~ obstime,
    subject = "patient",
    ng = 2,
    dynamic_covariates = c("CD4"),
    maxiter = 5000, rep = 25, nwg = TRUE
  ) |>
  predict_longitudinal(
    landmarks = c(6, 8),
    method = "lcmm",
    avg = TRUE,
    include_clusters = FALSE,
    var.time = "obstime",
    dynamic_covariates = c("CD4")
  ) |>
  fit_survival(
    formula = Surv(event_time, event_status) ~ drug,
    landmarks = c(6, 8),
    horizons = 12 + c(6, 8),
    method = "coxph",
    dynamic_covariates = c("CD4"),
    include_clusters = FALSE
  ) |>
  predict_survival(
    landmarks = c(6, 8),
    horizons = 12 + c(6, 8),
    dynamic_covariates = c("CD4"),
    include_clusters = FALSE
  )

## -----------------------------------------------------------------------------
summary(landmarking_object,
        type = "longitudinal",
        landmark = 6,
        dynamic_covariate = "CD4")

## -----------------------------------------------------------------------------
summary(landmarking_object, type = "survival", landmark = 6, horizon = 18)

## -----------------------------------------------------------------------------
performance_metrics(
  landmarking_object,
  landmarks = c(6, 8),
  horizons = c(18, 20),
  auc_t = TRUE, c_index = FALSE,
  h_times = c(3, 6, 12)
)