0. Setup

Setup

This is a repository with training material for Tumor Growth Inhibition (TGI) and joint TGI-OS (Overall Survival) modeling.

Here is an overview of the required setup steps, which are described in more detail below:

1. Install RTools (if you are on Windows)
2. Install necessary R packages
3. Install cmdstanr (optional but highly recommended)
4. Clone the repository from GitHub (https://github.com/RCONIS/tgi-os-training)
5. Open the folder in RStudio or VSCode

Install RTools

If you are running the examples on a Windows PC, you will need to install RTools, see
here.

You need to use the version of RTools that matches your R version. You can check your R version by running R.Version()$version.string.

You can then afterwards check the installation of RTools with:

if(!require(pkgbuild)) install.packages("pkgbuild")
pkgbuild::has_build_tools()

Install necessary R packages

The following code will install the packages that are required to run the examples in this repository.

install_if_not_available <- function(pkg, min_version = NULL) {
    is_installed <- suppressWarnings(suppressPackageStartupMessages(
      require(pkg, character.only = TRUE)
    ))
    if (is_installed & !is.null(min_version)) {
      version_ok <- packageVersion(pkg) >= min_version
    }    
    if (!is_installed | !version_ok) {
        install.packages(pkg)
    }
  }
packages <- c(
  "bayesplot", 
  "brms", 
  "ggplot2",
  "gt",
  "here", 
  "janitor",
  "modelr",
  "posterior",
  "readxl",
  "rstan",
  "tidybayes", 
  "tidyverse",   
  "truncnorm",
  "sn",
  "fuzzyjoin",
  "glue"
)
sapply(packages, install_if_not_available)
remotes::install_github("genentech/jmpost")

[Optional] Install cmdstanr

Optionally, you can use cmdstanr as the backend of brms for fitting the models.

There are a few advantages of using cmdstanr over the default rstan:

• With cmdstanr, you can cache the compiled model, therefore you don’t need to recompile the model as long as the same formula is used, even after restarting R session (or re-opening RStudio).
• cmdstanr is more up-to-date and more actively maintained compared to rstan.

A detailed installation guide for cmdstanr is available here. Here is a brief summary:

1. Install cmdstanr with:

   # Typically, you install cmdstanr from the R-universe as follows:
   install.packages("cmdstanr", repos = c('https://stan-dev.r-universe.dev', getOption("repos")))
   
   # If the above does not work, e.g. because you in a company network or VPN, you can install cmdstanr from GitHub instead:
   remotes::install_github("stan-dev/cmdstanr@*release")    

2. Check the compiler toolchain with:

   cmdstanr::check_cmdstan_toolchain()

   • This should not be a problem on Mac and Linux (including RStudio Cloud instances), but might be a problem on Windows.
   • If you have problems, please check the installation guide for Windows.
     • RTools is the easiest way.
     • Even if you have RTools, you might still see an error like "Rtools44 installation found but the toolchain was not installed.".
     • In this case, you can run cmdstanr::check_cmdstan_toolchain(fix = TRUE) and this will likely resolve the issue.

3. Install the CmdStan backend with:

   cmdstanr::install_cmdstan(cores = 2)

4. Testing the cmdstanr installation

   • If the above installation was successful, you should now be able to run the following simple model.

   library(cmdstanr)
   file <- file.path(
     cmdstan_path(), 
     "examples", "bernoulli", "bernoulli.stan"
   )
   mod <- cmdstan_model(file)
   data_list <- list(N = 10, y = c(0, 1, 0, 0, 0, 0, 0, 0, 0, 1))
   
   fit <- mod$sample(
     data = data_list,
     seed = 123,
     chains = 4,
     parallel_chains = 4,
     refresh = 500
   )