Question

我的数据看起来像这样，

posture code HR EE  a   b
cycling A03 102 100 3   6
standingA03 99  99  4   6
sitting A03 98  67  5   5
walking A03 97  78  3   6
cycling B01 111 76  5   5
standingB01 100 88  4   4
sitting B01 78  34  4   3
walking B01 99  99  2   2

我需要转置它，使它看起来如下：

code    cycling_HR  cycling_EE  cycling_a   cycling_b   standing_HR standing_EE standing_a  standing_b  sitting_HR  sitting_EE  sitting_a   sitting_b   walking_HR  walking_EE  walking_a   walking_b
A03     102    100  3       6   99          99          4   6   98  67  5   5   97  78  3   6
B01     111    76   5       5   100         88          4   4   78  34  4   3   99  99  2   2

等等（抱歉格式化）。我无法找到适当的答案来澄清问题。任何帮助都会受到欢迎。

Answer 1

这是一个非常基本的“长到宽”的重塑问题。

您可以使用reshape函数在基础R中执行此操作：

    reshape(mydf, direction = "wide", idvar = "code", timevar = "posture")
#   code HR.cycling EE.cycling a.cycling b.cycling HR.standing EE.standing
# 1  A03        102        100         3         6          99          99
# 5  B01        111         76         5         5         100          88
#   a.standing b.standing HR.sitting EE.sitting a.sitting b.sitting HR.walking
# 1          4          6         98         67         5         5         97
# 5          4          4         78         34         4         3         99
#   EE.walking a.walking b.walking
# 1         78         3         6
# 5         99         2         2

您还可以查看“dplyr”+“tidyr”方法，可能是这样的：

library(dplyr)
library(tidyr)
mydf %>%
  gather(var, val, HR:b) %>%
  unite(v1, posture, var) %>%
  spread(v1, val)

Answer 2

或者对于大数据集（因为reshape非常慢），您可以尝试data.table v>=1.9.5

library(data.table)
dcast(setDT(df), code ~ posture, value.var = c("HR", "EE", "a", "b"))
#    code cycling_HR sitting_HR standing_HR walking_HR cycling_EE sitting_EE standing_EE walking_EE cycling_a sitting_a standing_a walking_a cycling_b sitting_b
# 1:  A03        102         98          99         97        100         67          99         78         3         5          4         3         6         5
# 2:  B01        111         78         100         99         76         34          88         99         5         4          4         2         5         3
#    standing_b walking_b
# 1:          6         6
# 2:          4         2

略微更大的数据（400万行）的基准：

library(dplyr)
library(tidyr)
require(data.table)
set.seed(1L)
df = data.frame(posture = c("cycling", "standing", "sitting", "walking"), 
                code = rep(paste("A", 1:1e6, sep=""), each=4L), 
                HR = sample(120, 4e6, TRUE),
                EE = sample(100, 4e6, TRUE), 
                a = sample(5, 4e6, TRUE), 
                b = sample(10, 4e6, TRUE), 
                stringsAsFactors=FALSE)

# base R approach
system.time(reshape(df, direction = "wide", idvar = "code", timevar = "posture"))
#    user  system elapsed 
#  23.183   0.470  23.838 

# dplyr + tidyr
system.time({
df %>%
  gather(var, val, HR:b) %>%
  unite(v1, posture, var) %>%
  spread(v1, val)
})
#    user  system elapsed 
#  17.312   1.046  18.446 

# data.table
system.time(dcast(setDT(df), code ~ posture, 
            value.var = c("HR", "EE", "a", "b")))
#    user  system elapsed 
#   1.216   0.136   1.367

Answer 3

使用tidyr？

library(tidyr)

x<-data.frame(posture=c("cycling", "standing", "sitting", "walking"),
           code=c("A03", "A03", "B01", "B01"),
           HR=c(1,3,3,4),
           EE=c(1,3,3,5))

x2<-gather(x, key=type, value=vals, -c(code, posture))
x2$vars<-paste(x2$posture, x2$type, sep="_")

x2<-select(x2, -c(posture, type))
spread(x2, key=vars, value=vals)

在R中转置复杂的data.frame

3 个答案: