R中没有组的行的合并和部分添加答案

【问题标题】：Merger and partial addition of rows without groups in RR中没有组的行的合并和部分添加
【发布时间】：2018-10-09 22:42:08
【问题描述】：

以下是我为 dplyr 编写的问题的表述：

library(tidyverse)

df <- tibble(State = c("A", "A", "A", "A", "A", "A", "B", "B", "B"),
             District_code = c(1:9),
                 District = c("North", "West", "North West", "South", "East", "South East", 
                              "XYZ", "ZYX", "AGS"), 
                 Population = c(1000000, 2000000, 3000000, 4000000, 5000000, 6000000, 
                                7000000, 8000000, 9000000))

df
#> # A tibble: 9 x 4
#>   State District_code District   Population
#>   <chr>         <int> <chr>           <dbl>
#> 1 A                 1 North         1000000
#> 2 A                 2 West          2000000
#> 3 A                 3 North West    3000000
#> 4 A                 4 South         4000000
#> 5 A                 5 East          5000000
#> 6 A                 6 South East    6000000
#> 7 B                 7 XYZ           7000000
#> 8 B                 8 ZYX           8000000
#> 9 B                 9 AGS           9000000

对于某些州，我需要将使用名称的地区合并到更少的地理类别中。特别是，A 国应该只有：“North - West - North West”和“South - East - South East”。必须添加一些变量，例如人口；但其他像 District_code 应该获得 NA。我发现了this example 的跨行操作，但并不完全相同。 Grouping 似乎不适用。

最终的结果应该是这样的：

new_df
#> # A tibble: 5 x 4
#>   State District_code District                  Population
#>   <chr>         <int> <chr>                          <dbl>
#> 1 A                NA North - West - North West    5000000
#> 2 A                NA South - East - South East   15000000
#> 3 B                 7 XYZ                          7000000
#> 4 B                 8 ZYX                          8000000
#> 5 B                 9 AGS                          9000000

在实际数据框中，有许多变量（例如 Population）必须添加，还有一些其他变量（例如 District_code）必须获取 NA 值。

非常感谢您的帮助！

【问题讨论】：

标签： r dataframe dplyr

【解决方案1】：

您可以使用fct_collapse 指定新的因子水平，然后在新组上使用summarise。

df %>%
  mutate(District = 
           fct_collapse(District, 
                        "North - West - North West" = c("North", "West", "North West"), 
                        "South - East - South East" = c("South", "East", "South East"))) %>% 
  group_by(State, District) %>% 
  summarise(Population = sum(Population), 
            District_code = ifelse(n() > 1, NA_real_, District_code))

# A tibble: 5 x 3
# Groups:   State [?]
#   State District                  Population
#   <chr> <fct>                          <dbl>
# 1 A     South - East - South East   15000000
# 2 A     North - West - North West    6000000
# 3 B     AGS                          9000000
# 4 B     XYZ                          7000000
# 5 B     ZYX                          8000000

如果您只想更改某个特定州的地区，您可以像这样添加case_when 或if_else，并将汇总函数作为列类型的条件（这里是人口的两倍，而不是整数区）

df %>%
  mutate(District = 
           case_when(State == "A" ~ 
                       fct_collapse(District, 
                                    "North - West - North West" = c("North", "West", "North West"), 
                                    "South - East - South East" = c("South", "East", "South East")), 
                     TRUE ~ factor(District))) %>% 
  group_by(State, District) %>% 
  summarise_all(funs({if(is.double(.)) {
    sum(.) 
  } else {
    if (length(unique(.)) > 1) {
      NA
    } else {
      unique(.)
    }
  }}))

【讨论】：

谢谢凯丝！我认为您的第二个示例使其真正可扩展。如果我理解正确，只有那些是 dbls 的列被求和；当两个或更多已分组或保持其值时，其余部分要么变成 NA。正是我需要的！不过，我遇到了一个奇怪的错误：DISTRICT_CODE 列无法将第 138 组提升为字符。评估中的某些内容不太有效？
这很奇怪。 DISTRICT_CODE 列是否像您的示例中那样是整数？
DISTRICT_CODE 实际上是字符，我没有意识到这在您的解决方案中很重要；将其更改为 int 确实使整个事情运行...但是另一个字符变量 REGION_TYPE ["Urban", "Rural"] 只是全部设置为 NA？不太清楚这些事情是如何发生的，因为代码似乎无法区分不同类型的变量...... :-/ 感谢您提供任何进一步的线索！

【解决方案2】：

对于某些州，我需要将使用名称的地区合并到更少的地理类别中。特别是，A 国应该只有：“North - West - North West”和“South - East - South East”。

您需要写下分组规则，例如...

merge_rules = list(
  list(State = "A", District = c("North", "West", "North West")),
  list(State = "A", District = c("South", "East", "South East"))
)

必须添加一些变量，例如人口；但其他像 District_code 应该获得 NA。

我会通过将合并规则放在一个表格中来做到这一点；合并后进行计算；并对未合并的行进行 rbind-ing。这是data.table的方式...

library(data.table)
DT  = data.table(df)
mDT = rbindlist(lapply(merge_rules, as.data.table), id = "g")

gDT = DT[mDT, on=.(State, District)][, .(
  District_code = District_code[NA_integer_],
  District = paste(District, collapse = " - "),
  Population = sum(Population)
), by=.(g, State)]

rbind(
  DT[!mDT, on=.(State, District)],
  gDT[, !"g"]
)[order(State, District)]

   State District_code                  District Population
1:     A            NA North - West - North West    6.0e+06
2:     A            NA South - East - South East    1.5e+07
3:     B             9                       AGS    9.0e+06
4:     B             7                       XYZ    7.0e+06
5:     B             8                       ZYX    8.0e+06

而且，我猜，tidyverse 的方式是相似的：

mtib = bind_rows(lapply(merge_rules, as.tibble), .id = "g")

gtib = right_join(df, mtib, by=c("State", "District")) %>% 
  group_by(g, State) %>% summarise(
    District_code = District_code[NA_integer_],
    District = paste(District, collapse = " - "),
    Population = sum(Population)    
  )

bind_rows(
  anti_join(df, mtib, by=c("State", "District")),
  gtib %>% ungroup %>% select(-g)
) %>% arrange(State, District)

# A tibble: 5 x 4
  State District_code District                  Population
  <chr>         <int> <chr>                          <dbl>
1 A                NA North - West - North West    6000000
2 A                NA South - East - South East   15000000
3 B                 9 AGS                          9000000
4 B                 7 XYZ                          7000000
5 B                 8 ZYX                          8000000

【讨论】：

【解决方案3】：

这是获取状态 A 的聚合人口的一种方法：

df %>% 
  filter(State == "A") %>%
  mutate(`North - West - North West` = (District == "North"|District == "West"|District == "North West"), 
         `South - East - South East` = (District == "South"|District == "East"|District == "South East")) %>% 
  gather(key = Districts, value = present, 5:6) %>% 
  filter(present != FALSE) %>% 
  group_by(Districts) %>% 
  summarise(Population = sum(Population))

它给出了输出：

  Districts          Population
  <chr>                   <dbl>
1 North - West - No…    6000000
2 South - East - So…   15000000

应该有人能帮我们把上面的东西放到原来的df中。

【讨论】：