希望优化 R 中的 mutate(case_when( )) function，在系统命名的变量列表中需要相同的突变

Question

我试图在 R 中找到一种矢量化的通用方法来完成我通常使用 Stata 的任务。 我有一个具有共同结构的变量列表，我希望有效地对所有变量应用相同的转换。 每个变量有两个版本（x 和 m），如果观察分别位于第 1 组或第 2 组，则每个变量只有一个值（一些是数字，一些是字符）。 我正在尝试创建一个通用变量，它只包含该行的相应非 NA 值。 在 Stata 中，我将使用以下代码完成此操作：

foreach word in cost cost_add25 cost_less25 parkingaccess parkingaccess_add25 parkingaccess_less25 timeresult timeresult_add25 timeresult_less25 timetravel timetravel_add25 timetravel_less25 timewait timewait_add25 timewait_less25 freeparking onlinesch service service_less1 service_plus1 pcprec specialty {
    gen statusquo_`word' = statusquo_x_`word' if group == 1
    replace statusquo_`word' = statusquo_m_`word' if group == 2
}

然而，到目前为止，我一直对如何在 R 中使用矢量化函数有效地执行此操作感到困惑——我尝试过使用 purrr:map、lapply 和 mapply，但效果不佳。 这是我当前的实现，非常粗糙：

database <- database %>%
  mutate(
    statusquo_cost                 = case_when(group == 1 ~ statusquo_x_cost                 , group == 2 ~ statusquo_m_cost                 ),
    statusquo_cost_add25           = case_when(group == 1 ~ statusquo_x_cost_add25           , group == 2 ~ statusquo_m_cost_add25           ),
    statusquo_cost_less25          = case_when(group == 1 ~ statusquo_x_cost_less25          , group == 2 ~ statusquo_m_cost_less25          ),
    statusquo_parkingaccess        = case_when(group == 1 ~ statusquo_x_parkingaccess        , group == 2 ~ statusquo_m_parkingaccess        ),
    statusquo_parkingaccess_add25  = case_when(group == 1 ~ statusquo_x_parkingaccess_add25  , group == 2 ~ statusquo_m_parkingaccess_add25  ),
    statusquo_parkingaccess_less25 = case_when(group == 1 ~ statusquo_x_parkingaccess_less25 , group == 2 ~ statusquo_m_parkingaccess_less25 ),
    statusquo_timeresult           = case_when(group == 1 ~ statusquo_x_timeresult           , group == 2 ~ statusquo_m_timeresult           ),
    statusquo_timeresult_add25     = case_when(group == 1 ~ statusquo_x_timeresult_add25     , group == 2 ~ statusquo_m_timeresult_add25     ),
    statusquo_timeresult_less25    = case_when(group == 1 ~ statusquo_x_timeresult_less25    , group == 2 ~ statusquo_m_timeresult_less25    ),
    statusquo_timetravel           = case_when(group == 1 ~ statusquo_x_timetravel           , group == 2 ~ statusquo_m_timetravel           ),
    statusquo_timetravel_add25     = case_when(group == 1 ~ statusquo_x_timetravel_add25     , group == 2 ~ statusquo_m_timetravel_add25     ),
    statusquo_timetravel_less25    = case_when(group == 1 ~ statusquo_x_timetravel_less25    , group == 2 ~ statusquo_m_timetravel_less25    ),
    statusquo_timewait             = case_when(group == 1 ~ statusquo_x_timewait             , group == 2 ~ statusquo_m_timewait             ),
    statusquo_timewait_add25       = case_when(group == 1 ~ statusquo_x_timewait_add25       , group == 2 ~ statusquo_m_timewait_add25       ),
    statusquo_timewait_less25      = case_when(group == 1 ~ statusquo_x_timewait_less25      , group == 2 ~ statusquo_m_timewait_less25      ),
    statusquo_freeparking          = case_when(group == 1 ~ statusquo_x_freeparking          , group == 2 ~ statusquo_m_freeparking          ),
    statusquo_onlinesch            = case_when(group == 1 ~ statusquo_x_onlinesch            , group == 2 ~ statusquo_m_onlinesch            ),
    statusquo_service              = case_when(group == 1 ~ statusquo_x_service              , group == 2 ~ statusquo_m_service              ),
    statusquo_service_less1        = case_when(group == 1 ~ statusquo_x_service_less1        , group == 2 ~ statusquo_m_service_less1        ),
    statusquo_service_plus1        = case_when(group == 1 ~ statusquo_x_service_plus1        , group == 2 ~ statusquo_m_service_plus1        ),
    statusquo_pcprec               = case_when(group == 1 ~ statusquo_x_pcprec               , group == 2 ~ statusquo_m_pcprec               ),
    statusquo_specialty            = case_when(group == 1 ~ statusquo_x_specialty            , group == 2 ~ statusquo_m_specialty            )
    )   

在一个理想的世界中，我会有这样的东西，其中“VAR”在上面列表中的值之间循环，但是 R 对类型非常敏感，并将新变量名称视为文本。

vars <- c("cost","cost_add25","cost_less25","parkingaccess","parkingaccess_add25","parkingaccess_less25","timeresult","timeresult_add25","timeresult_less25","timetravel","timetravel_add25","timetravel_less25","timewait","timewait_add25","timewait_less25","freeparking","onlinesch","service","service_less1","service_plus1","pcprec","specialty")

database <- database %>%
  mutate(
    statusquo_VAR = case_when(group == 1 ~ statusquo_x_VAR, group == 2 ~ statusquo_m_VAR)
  )  

任何建议都会非常有帮助，如果需要，我可以提供数据样本。 尽管我不确定这个问题是否复杂到需要一个。

我尝试使用矢量化操作将两个变量的列表组合在一起，基于它们名称的结构化列表，如果没有繁琐的方法我无法做到这一点。

Answer 1

我已经修改了你的代码，希望它能得到你想要的。 另一个小问题，我用ifelse代替case_when ，但两者都应该有效。

vars <- c("cost","cost_add25","cost_less25","parkingaccess","parkingaccess_add25","parkingaccess_less25","timeresult","timeresult_add25","timeresult_less25","timetravel","timetravel_add25","timetravel_less25","timewait","timewait_add25","timewait_less25","freeparking","onlinesch","service","service_less1","service_plus1","pcprec","specialty")

for (i in seq_along(vars)){
database <- database %>%
  mutate(
    !!paste0("statusquo_", vars[i]) := ifelse(group == 1, get(paste0("statusquo_x_", vars[i])), get(paste0("statusquo_m_", vars[i])))
  ) 
}

编辑：

使用较新版本的 dplyr >= 1.0

for (i in vars){
  database <- database %>%
    mutate(
      "statusquo_{i}" := ifelse(group == 1, get(paste0("statusquo_x_", i)), get(paste0("statusquo_m_", i)))
    ) 
}