如何使用Dplyr的Summarize和which（）查找最小值/最大值

Question

我有以下數據：

Name <- c("Sam", "Sarah", "Jim", "Fred", "James", "Sally", "Andrew", "John", "Mairin", "Kate", "Sasha", "Ray", "Ed")
Age <- c(22,12,31,35,58,82,17,34,12,24,44,67,43)
Group <- c("A", "B", "B", "B", "B", "C", "C", "D", "D", "D", "D", "D", "D") 
data <- data.frame(Name, Age, Group)

我想用dplyr

（1）按“組”將數據分組（2）顯示每個組中的最小和最大年齡（3）顯示具有最小和最大年齡的人的姓名

下面的代碼執行此操作：

data %>% group_by(Group) %>%
     summarize(minAge = min(Age), minAgeName = Name[which(Age == min(Age))], 
               maxAge = max(Age), maxAgeName = Name[which(Age == max(Age))])

哪個效果很好：

  Group minAge minAgeName maxAge maxAgeName
1     A     22        Sam     22        Sam
2     B     12      Sarah     58      James
3     C     17     Andrew     82      Sally
4     D     12     Mairin     67        Ray

但是，如果有多個最小值或最大值，我就會遇到問題：

Name <- c("Sam", "Sarah", "Jim", "Fred", "James", "Sally", "Andrew", "John", "Mairin", "Kate", "Sasha", "Ray", "Ed")
Age <- c(22,31,31,35,58,82,17,34,12,24,44,67,43)
Group <- c("A", "B", "B", "B", "B", "C", "C", "D", "D", "D", "D", "D", "D") 
data <- data.frame(Name, Age, Group)

> data %>% group_by(Group) %>%
+   summarize(minAge = min(Age), minAgeName = Name[which(Age == min(Age))], 
+             maxAge = max(Age), maxAgeName = Name[which(Age == max(Age))])
Error: expecting a single value

我正在尋找兩種解決方案：

（1）顯示哪個最小或最大名稱無關緊要，僅顯示一個（即找到的第一個值）（2）如果存在“聯系”，則顯示所有最小值和最大值

如果不清楚，請讓我知道，並提前致謝！

Answer 1

您可以使用which.min和which.max獲得第一個值。

data %>% group_by(Group) %>%
  summarize(minAge = min(Age), minAgeName = Name[which.min(Age)], 
            maxAge = max(Age), maxAgeName = Name[which.max(Age)])

要獲取所有值，請使用例如粘貼適當的collapse參數。

data %>% group_by(Group) %>%
  summarize(minAge = min(Age), minAgeName = paste(Name[which(Age == min(Age))], collapse = ", "), 
            maxAge = max(Age), maxAgeName = paste(Name[which(Age == max(Age))], collapse = ", "))

Answer 2

我實際上建議您將數據保留為“長”格式。 這是我的處理方法：

library(dplyr)

有聯系時保留所有值：

data %>%
  group_by(Group) %>%
  arrange(Age) %>%  ## optional
  filter(Age %in% range(Age))
# Source: local data frame [8 x 3]
# Groups: Group
# 
#     Name Age Group
# 1    Sam  22     A
# 2  Sarah  31     B
# 3    Jim  31     B
# 4  James  58     B
# 5 Andrew  17     C
# 6  Sally  82     C
# 7 Mairin  12     D
# 8    Ray  67     D

有聯系時僅保留一個值：

data %>%
  group_by(Group) %>%
  arrange(Age) %>%
  slice(if (length(Age) == 1) 1 else c(1, n())) ## maybe overkill?
# Source: local data frame [7 x 3]
# Groups: Group
# 
#     Name Age Group
# 1    Sam  22     A
# 2  Sarah  31     B
# 3  James  58     B
# 4 Andrew  17     C
# 5  Sally  82     C
# 6 Mairin  12     D
# 7    Ray  67     D

---

如果您真的想要一個“寬”數據集，則基本概念是使用“ tidyr”來gather和spread數據：

library(dplyr)
library(tidyr)

data %>%
  group_by(Group) %>%
  arrange(Age) %>%
  slice(c(1, n())) %>%
  mutate(minmax = c("min", "max")) %>%
  gather(var, val, Name:Age) %>%
  unite(key, minmax, var) %>%
  spread(key, val)
# Source: local data frame [4 x 5]
# 
#   Group max_Age max_Name min_Age min_Name
# 1     A      22      Sam      22      Sam
# 2     B      58    James      31    Sarah
# 3     C      82    Sally      17   Andrew
# 4     D      67      Ray      12   Mairin

盡管您想要聯系的廣泛形式還不清楚。

Answer 3

這是一些data.table方法，第一個是從@akrun借來的：

setDT(data)

# show one, wide format
data[,c(min=.SD[which.min(Age)],max=.SD[which.max(Age)]),by=Group]
   # Group min.Name min.Age max.Name max.Age
# 1:     A      Sam      22      Sam      22
# 2:     B    Sarah      31    James      58
# 3:     C   Andrew      17    Sally      82
# 4:     D   Mairin      12      Ray      67

# show all, long format
data[,{
  mina=min(Age)
  maxa=max(Age)
  rbind(
    data.table(minmax="min",Age=mina,Name=Name[which(Age==mina)]),
    data.table(minmax="max",Age=maxa,Name=Name[which(Age==maxa)])
)},by=Group]
   # Group minmax Age   Name
# 1:     A    min  22    Sam
# 2:     A    max  22    Sam
# 3:     B    min  31  Sarah
# 4:     B    min  31    Jim
# 5:     B    max  58  James
# 6:     C    min  17 Andrew
# 7:     C    max  82  Sally
# 8:     D    min  12 Mairin
# 9:     D    max  67    Ray    

我認為長格式是最好的，因為它允許您使用minmax進行過濾，但是代碼很折磨且效率低下。

這里有一些不太好的方法：

# show all, wide format (with a list column)
data[,{
  mina=min(Age)
  maxa=max(Age)
  list(
    minAge=mina,
    maxAge=maxa,
    minNames=list(Name[Age==mina]),
    maxNames=list(Name[Age==maxa]))
},by=Group]
   # Group minAge maxAge  minNames maxNames
# 1:     A     22     22       Sam      Sam
# 2:     B     31     58 Sarah,Jim    James
# 3:     C     17     82    Andrew    Sally
# 4:     D     12     67    Mairin      Ray


# show all, wide format (with a string column)
# (just look at @shadow's answer)