I have a question for the parallel computing in R. I am going to use parallel computing to send each of the four bootstrap standard error computations (one for each response: Total_lignin, Glucose, Xylose and arabinose) to a different core and also report estimates of each maximum with corresponding standard errors. The concentration is a predictor.
I have already created a function called seBootFun that takes in response, predictor, B, and data and returns the standard deviation of the bootstrapped estimates and it works out when treating Glucose as a response but it fails when I used parallel computing. Could you please help me with it? The error is shown.
# Load the libraries
library(dplyr)
library(tidyverse)
# Read the .csv and only use M.giganteus and S.ravennae.
dat <- read_csv('concentration.csv') %>%
filter(variety == 'M.giganteus' | variety == 'S.ravennae') %>%
arrange(variety)
# Set seed for reproducibility purpose
set.seed(12)
# Sample size
n <- nrow(dat)
# Create the function to return the sd of the estimates
seBootFun <- function(resp, pred, B, data){
# A function for return the max value of predictor
max <- function(resp, pred, data){
# Draw the sample size from the dataset
sample <- slice_sample(.data = dat, n, prop = 1, replace = TRUE)
# A quadratic model fit
y <- as.matrix(sample[, resp])
x <- as.matrix(sample[, pred])
fit <- lm(y ~ x + I(x^2))
# Derive the max of the value of concentration
max <- -fit$coefficients[2]/(2*fit$coefficients[3])
return(max)
}
maxs <- replicate(B, max(resp, pred,
data = dat))
return(c(mean(maxs), sd(maxs)))
}
# Output
result <- seBootFun(resp = 'Glucose', pred = 'concentration', B = 5000, data = dat)
names(result) <- c('estimated', 'sd')
result
# Load the `parallel` library
library(parallel)
# Set seed and iterated times
set.seed(8)
# Set up cores
cores <- detectCores()
cluster <- makeCluster(cores - 1)
clusterExport(cluster, list("seBootFun", "max", "maxs"))
clusterEvalQ(cluster, library(tidyverse))
clusterEvalQ(cluster, library(dplyr))
result.1 <-
parLapply(cluster, fun = seBootFun(resp = 'Total_lignin', pred = 'concentration', B = 5000, data = dat))
I have already created a function called seBootFun that takes in response, predictor, B, and data and returns the standard deviation of the bootstrapped estimates and it works out when treating Glucose as a response but it fails when I used parallel computing. Could you please help me with it? The error is shown.
# Load the libraries
library(dplyr)
library(tidyverse)
# Read the .csv and only use M.giganteus and S.ravennae.
dat <- read_csv('concentration.csv') %>%
filter(variety == 'M.giganteus' | variety == 'S.ravennae') %>%
arrange(variety)
# Set seed for reproducibility purpose
set.seed(12)
# Sample size
n <- nrow(dat)
# Create the function to return the sd of the estimates
seBootFun <- function(resp, pred, B, data){
# A function for return the max value of predictor
max <- function(resp, pred, data){
# Draw the sample size from the dataset
sample <- slice_sample(.data = dat, n, prop = 1, replace = TRUE)
# A quadratic model fit
y <- as.matrix(sample[, resp])
x <- as.matrix(sample[, pred])
fit <- lm(y ~ x + I(x^2))
# Derive the max of the value of concentration
max <- -fit$coefficients[2]/(2*fit$coefficients[3])
return(max)
}
maxs <- replicate(B, max(resp, pred,
data = dat))
return(c(mean(maxs), sd(maxs)))
}
# Output
result <- seBootFun(resp = 'Glucose', pred = 'concentration', B = 5000, data = dat)
names(result) <- c('estimated', 'sd')
result
# Load the `parallel` library
library(parallel)
# Set seed and iterated times
set.seed(8)
# Set up cores
cores <- detectCores()
cluster <- makeCluster(cores - 1)
clusterExport(cluster, list("seBootFun", "max", "maxs"))
clusterEvalQ(cluster, library(tidyverse))
clusterEvalQ(cluster, library(dplyr))
result.1 <-
parLapply(cluster, fun = seBootFun(resp = 'Total_lignin', pred = 'concentration', B = 5000, data = dat))
