R调查问卷中的多核论点

Question

When attempting to use the multicore = TRUE argument in R's survey package, I get the following error:

Error in glm.control(multicore = TRUE) :
  unused argument (multicore = TRUE)

Example code for replication:

library(tidyverse)
library(survey)

mtcars %<>%
  mutate(weights = 1) %>%
  mutate(id = ceiling(row_number()/2)) %>%
  mutate(strata = ceiling(row_number()/4))

design <- svydesign(id =~ id, weights =~ weights, strata =~ strata,
                data = mtcars)

model <- svyglm(design = design, mpg ~ drat + wt + qsec,
            multicore=TRUE)

mutlicore does appear as a possible argument in the documentation for svyglm . This thread indicates that this argument may not work on Windows, but I'm running this on a Unix system.

I know the multicore package has been deprecated in favor of parallel , and the documentation here suggests that the survey package has been updated to reflect this change (see revision 3.29-5), but I'm not sure if or how that may be involved in this issue.

I've also tried installing and loading the obsolete multicore with the following code, but this does fix the issue.

url <- "https://cran.r-project.org/src/contrib/Archive/multicore/multicore_0.2.tar.gz"
pkgFile <- "multicore_0.2.tar.gz"
download.file(url, destfile = pkgFile)
install.packages(pkgs = pkgFile, type="source", repos=NULL)
library(multicore)

Answer 1

multicore =选项仅在复制权重设计和svyby（）中实现，因为它们是唯一允许多核支持的粗粒度并行性的分析。