Question

我尝试对不同的数据集进行一些回归。

加载库+数据

libraries = c('AER','plm','dplyr','lmtest','foreach','doParallel')
lapply(libraries, require, character.only = TRUE)
no_cores <- detectCores() - 8
cl <- makeCluster(no_cores, type="PSOCK")
registerDoParallel(cl)
data("Fatalities")
df1 = Fatalities
df2 = Fatalities %>% filter(spirits>1.3)

独立回归 首先，我在循环外运行回归以测试动态变量引用，并且它按预期工作。

yvar= as.symbol('fatal')
xvars =  rlang::parse_expr("beertax + baptist + income + emppop")
data_to_use = as.symbol("df1")

fatal_fe_mod <- eval(bquote(plm( .(yvar) ~ .(xvars), data = .(data_to_use), index = c("state", "year"),  model = "within", effect= "time")))
summary(fatal_fe_mod)
mod_summary = eval(bquote(coeftest(fatal_fe_mod, vcov=vcovHC(fatal_fe_mod,type="HC0",cluster="group"))))  # clustered errors
mod_summary

常规循环 接下来，我在常规的for循环中运行回归，该回归同样有效。

datasets = c("df1", "df2")
for(num_spec in c(1:2))  {
  print(num_spec)
  yvar= as.symbol('fatal')
  xvars =  rlang::parse_expr("beertax + baptist + income + emppop")
  data_to_use = as.symbol(datasets[num_spec])
  fatal_fe_mod <- eval(bquote(plm( .(yvar) ~ .(xvars), data = .(data_to_use), index = c("state", "year"),  model = "within", effect= "time")))
  summary(fatal_fe_mod)
  mod_summary = eval(bquote(coeftest(fatal_fe_mod, vcov=vcovHC(fatal_fe_mod,type="HC0",cluster="group"))))  # clustered errors
  print(mod_summary)

}

并行循环 不幸的是，并行循环失败。它立即返回错误。看来问题与.(data_to_use)步骤有关。在并行循环内部，R似乎无法访问数据帧。

foreach(num_spec=1:2, .packages=libraries) %dopar% {
  yvar= as.symbol('fatal')
  xvars =  rlang::parse_expr("beertax + baptist + income + emppop")
  data_to_use = as.symbol(datasets[num_spec])
  fatal_fe_mod <- eval(bquote(plm( .(yvar) ~ .(xvars), data = .(data_to_use), index = c("state", "year"),  model = "within", effect= "time")))
  summary(fatal_fe_mod)
  mod_summary = eval(bquote(coeftest(fatal_fe_mod, vcov=vcovHC(fatal_fe_mod,type="HC0",cluster="group"))))  # clustered errors
  mod_summary

}

最后一个立即返回：

Error in { : task 1 failed - "object 'df1' not found"

R动态变量引用在并行循环中失败（但在常规循环中有效）

0 个答案: