vignettes/bhmbasket_on_HPC.R

In bhmbasket: Bayesian Hierarchical Models for Basket Trials

## ----knitr_options, include = FALSE-------------------------------------------
knitr::opts_chunk$set(
  eval     = FALSE,  # en- / disables R code evaluation globally
  cache    = FALSE,  # en- / disables R code caching globally
  collapse = TRUE,
  comment  = "#>"
)


## ----setup--------------------------------------------------------------------
library(bhmbasket)
library(doFuture)
library(future.batchtools)

rng_seed <- 5440
set.seed(rng_seed)


## ----SLURM_Setup--------------------------------------------------------------
registerDoFuture()

job_time  <- 24  # time for job in hours
n_workers <- 10  # number of worker nodes
n_cpus    <- 16  # number of cpus per worker node
gb_memory <- 1   # memory [GB] per worker node

slurm <- tweak(batchtools_slurm,
           template  = system.file('templates/slurm-simple.tmpl',
                                   package = 'batchtools'),
           workers   = n_workers,
           resources = list(
             walltime  = 60 * 60 * job_time,
             ncpus     = n_cpus,
             memory    = 1000 * gb_memory))

plan(slurm)


## -----------------------------------------------------------------------------
scenarios_list <- simulateScenarios(
    n_subjects_list     = list(c(10, 20, 30)),
    response_rates_list = list(c(0.1, 0.2, 3)),
    n_trials            = 10)

  analyses_list <- performAnalyses(
    scenario_list       = scenarios_list,
    target_rates        = c(0.1, 0.1, 0.1),
    calc_differences    = matrix(c(3, 2, 2, 1), ncol = 2),
    n_mcmc_iterations   = 100)

  getEstimates(analyses_list)