dplyr和interval:不使用循环计算观察值和求和数据

Question

我想使用R dplyr包来计算以下间隔相关问题而不使用循环:

1. 我想计算每个区间的观察结果(绝对和相对区间终点)
2. 我想在每个区间(绝对和相对区间端点)中总结观察数据

间隔端点来自列df_abs $ interval和df_rel $ interval.例如

1. 间隔:( - inf,-60]
2. 间隔:( - 60,-30]
3. 间隔:( - 30,0)

具有数据和间隔的数据框如下所示:

library(dplyr)

# ----------{ data and interval ----------
df_data <- data.frame(varA = NA,
                      varB = NA, 
                      varC = c(-81.0, -14.3, 29.6, 42.7, 46.4, 57.7, 15.3, 256.3,  20.3, -25.1, -23.1, -17.5))

df_abs <- data.frame(interval = c(-Inf, -60, -30, 0, 30, 60, 100, 200, Inf),
                     count = NA,
                     sum = NA)

df_rel <- data.frame(interval = c(0,5,15,50,75,95,100),
                     count = NA,
                     sum = NA)
# ---------- data and interval }----------


# ----------{ calculation ----------    
# absolute data frame
for (i in 1 : nrow(df_abs)-1) {  
  # count observation between interval
  df_abs$count[i+1]  <- summarise(df_data, sum(df_abs$interval[i] < varC &   varC <= df_abs$interval[i+1]))

  # sum between interval
  df_abs$sum[i+1]  <- sum(df_data$varC[df_abs$interval[i] < df_data$varC &   df_data$varC <= df_abs$interval[i+1]])  
}


# relative data frame
df_data_arranged <- df_data %>%
                      arrange(varC) %>%
                      mutate(observationPercent = c(1:nrow(df_data)) * 100/length(df_data$varC))


for (i in 1 : nrow(df_rel)-1) {  
  # count observation between interval
  df_rel$count[i+1]  <- summarise(df_data_arranged, sum(df_rel$interval[i] < observationPercent &   observationPercent <= df_rel$interval[i+1])) 

  # sum between interval
  df_rel$sum[i+1]  <- sum(df_data_arranged$varC[df_rel$interval[i] < df_data_arranged$observationPercent & df_data_arranged$observationPercent <= df_rel$interval[i+1]]) 
}    
# ---------- calculation }----------

答案应如下所示:

df_abs <- data.frame(interval = c(-Inf, -60, -30, 0, 30, 60, 100, 200, Inf),
                     count = c(0,1,0,4,3,3,0,0,1),
                     sum = c(0,-81,0,-80,65.2,146.8,0,0,256.3))    

df_rel <- data.frame(interval = c(0,5,15,50,75,95,100),
                     count = c(0,0,1,4,3,2,1),
                     sum = c(0,0,-81,-39.6,92.6,104.1,256.3))

据我了解dplyr软件包,对于这两个问题中的每一个都应该有一个相当简短的直接解决方案,而不必使用循环.

Answer 1

这可以按如下方式完成：

• 创建一个新列 ( mutate) 来标识哪个观测值属于哪个区间（通过base::cut）

• group_by按间隔 ( )对观察结果进行分组

• 对结果应用您的操作（此处summarise使用dplyr'sn()和 a common ）sum

如下：

df_abs <- mutate(df_data, interval = cut(varC, df_abs$interval)) %>% 
  group_by(interval) %>% 
  summarise(count=n(), sum=sum(varC))
#     interval count   sum
#1 (-Inf,-60]     1 -81.0
#2    (-30,0]     4 -80.0
#3     (0,30]     3  65.2
#4    (30,60]     3 146.8
#5 (200, Inf]     1 256.3

df_rel <- mutate(df_data_arranged, 
                 interval = cut(observationPercent, df_rel$interval)) %>% 
  group_by(interval) %>% 
  summarise(count=n(), sum=sum(varC))
#  interval count   sum
#1   (5,15]     1 -81.0
#2  (15,50]     5 -64.7
#3  (50,75]     3  92.6
#4  (75,95]     2 104.1
#5 (95,100]     1 256.3