sandbox/00-makedata.R

In benkeser/haltmle.sim: Simulation study for HAL-TMLE

#! /usr/bin/env Rscript 
# This file was used to submit the simulation files to 
# a slurm-based Unix system. Using the sce.sh shell script
# one can submit each simulation in sequence. First, data files are
# created for each simulation. Those data files are then analyzed 
# in the 'run' execution. Then the results are collated in the 'merge'
# execution. 

# get environment variables
MYSCRATCH <- Sys.getenv('MYSCRATCH')
RESULTDIR <- Sys.getenv('RESULTDIR')
STEPSIZE <- as.numeric(Sys.getenv('STEPSIZE'))
TASKID <- as.numeric(Sys.getenv('SLURM_ARRAY_TASK_ID'))

# set defaults if nothing comes from environment variables
MYSCRATCH[is.na(MYSCRATCH)] <- '.'
RESULTDIR[is.na(RESULTDIR)] <- '.'
STEPSIZE[is.na(STEPSIZE)] <- 1
TASKID[is.na(TASKID)] <- 0

# get command lines arguments
args <- commandArgs(trailingOnly = TRUE)
if(length(args) < 1){
  stop("Not enough arguments. Please use args 'listsize', 'prepare', 'run <itemsize>' or 'merge'")
}

# # load packages
library(arm)
library(plyr)
library(gam, lib.loc = "/home/dbenkese/R/x86_64-unknown-linux-gnu-library/3.2")
library(caret)
library(haltmle.sim, lib.loc = "/home/dbenkese/R/x86_64-unknown-linux-gnu-library/3.2")
library(Rsolnp, lib.loc = "/home/dbenkese/R/x86_64-unknown-linux-gnu-library/3.2")
library(future, lib.loc = "/home/dbenkese/R/x86_64-unknown-linux-gnu-library/3.2")
library(cvma, lib.loc = "/home/dbenkese/R/x86_64-unknown-linux-gnu-library/3.2")
library(hal9001, lib.loc = "/home/dbenkese/R/x86_64-unknown-linux-gnu-library/3.2")
library(drtmle, lib.loc = "/home/dbenkese/R/x86_64-unknown-linux-gnu-library/3.2")
library(SuperLearner)

ns <- c(200, 2000)
min_R2 <- c(0.01, seq(0.101, 0.901, by = 0.1))
max_R2 <- c(seq(0.1,0.9,0.1),0.99)
mat_R2 <- cbind(min_R2, max_R2)
bigB <- 1000

# # simulation parameters
parm <- expand.grid(seed=1:bigB,
                    n=ns,
                    range_R2 = split(mat_R2, row(mat_R2)))

# directories to save in 
saveDir <- "~/haltmle.sim/out/"
scratchDir <- "~/haltmle.sim/scratch/"

# get the list size #########
if (args[1] == 'listsize') {
  cat(nrow(parm))
}

# execute prepare job ##################
if (args[1] == 'prepare') {
  
}

# execute parallel job #################################################
if (args[1] == 'run') {
  if (length(args) < 2) {
    stop("Not enough arguments. 'run' needs a second argument 'id'")
  }
  id <- as.numeric(args[2])
  print(paste(Sys.time(), "arrid:" , id, "TASKID:",
              TASKID, "STEPSIZE:", STEPSIZE))
  for (i in (id+TASKID):(id+TASKID+STEPSIZE-1)) {
    print(paste(Sys.time(), "i:" , i))
    print(parm[i,])

    set.seed(parm$seed[i])
    dat <- haltmle.sim:::makeRandomDataT(n=parm$n[i], maxD = 8,
                                        minR2 = parm$range_R2[[i]][1],
                                        maxR2 = parm$range_R2[[i]][2])
    folds <- rep(seq_len(10), length = parm$n[i])
    folds <- sample(folds)
  
    save(list(data = dat, folds = folds), file=paste0(scratchDir,"draw_n=",parm$n[i],"_seed=",parm$seed[i],
                          "_r2=",parm$range_R2[[i]][1],".RData"))
}

# merge job ###########################
if (args[1] == 'merge') {
  
}