tidyverse pivot_longer 几组列，但避免中间 mutate_wider 步骤 [英] tidyverse pivot_longer several sets of columns, but avoid intermediate mutate_wider steps

问题描述

我有以下数据

dat <- data.frame(id         = c("A", "B", "C"),
                  Q1r1_pepsi = c(1,0,1),
                  Q1r1_cola  = c(0,0,1),
                  Q1r2_pepsi = c(1,1,1),
                  Q1r2_cola  = c(0,1,1),
                  stringsAsFactors = FALSE)

其中 Q1r1 和 Q1r2 是调查中的评级问题，百事可乐和可乐是被评级的品牌.所以我对两个品牌(百事可乐、可乐)有两个评分(r1 和 r2):

where Q1r1 and Q1r2 are rating questions in a survey and pepsi and cola are the brands being rated. So I have two ratings (r1 and r2) for two brands (pepsi, cola):

id      Q1r1_c1    Q1r1_c2    Q1r2_c1    Q1r2_c2
"A"     1          0          1          0
"B"     0          0          1          1
"C"     1          1          1          1

(附带问题:如何格式化 SO 帖子，以便它正确包含在 R 控制台中调用 dat 时获得的格式良好的输出?)

(Side question: how do I format a SO post so that it correctly contains the nicely formatted output that I would get when calling dat in the R Console?)

为了分析数据，我需要重新调整(旋转)数据，以便行表示唯一的评级-品牌对.因此，预期的结果是:

To analyze the data I need to reshape (pivot) the data so that rows indicate unique rating-brand pairs. Thus, the expected outcome would be:

id      brand   Q1r1    Q1r2
"A"     "pepsi" 1       1
"A"     "cola"  0       0
"B"     "pepsi" 0       1
"B"     "cola"  0       1
"C"     "pepsi" 1       1
"C"     "cola"  1       1

目前，我总是结合使用 pivot_longer 和 pivot_wider，但我希望我可以通过 pivoting_longer 直接得到这个结果，而无需做中间步骤:

Currently, I always do a combination of pivot_longer and pivot_wider, but I was hoping that I can directly get this result by pivoting_longer without doing the intermediate step:

library(tidyverse)

dat_long <- dat %>%
  pivot_longer(cols = starts_with("Q1")) %>%
  separate(name, into = c("item", "brand"), remove = FALSE)

dat_wide <- dat_long %>%
  pivot_wider(id_cols = c(id, brand),
              names_from = item,
              values_from = value)

在当前的示例中，执行此中间步骤仍然可以，但在其他不太干净的示例中会变得令人厌烦，例如假设我的列没有以 Q1r1_c1, Q1r1_c2, Q1r2_c1, Q1r2_c2 的良好结构命名，而是 Q4, Q5, Q8r1, Q8r2 地图所在的位置分别在 Q4 和 Q8r1 之间，以及 Q5/Q8r2 之间.

With this current example it's still ok to do this intermediate step, but it gets tiresome in other less clean examples, e.g. suppose my columns weren't named in a nice structure with Q1r1_c1, Q1r1_c2, Q1r2_c1, Q1r2_c2, but instead would be Q4, Q5, Q8r1, Q8r2 where the map would be between Q4 and Q8r1, and Q5/Q8r2, respectively.

推荐答案

您可以使用:

tidyr::pivot_longer(dat, cols = -id, 
                   names_to = c('.value', 'brand'), 
                   names_sep = "_")


#  id    brand  Q1r1  Q1r2
#  <chr> <chr> <dbl> <dbl>
#1 A     pepsi     1     1
#2 A     cola      0     0
#3 B     pepsi     0     1
#4 B     cola      0     1
#5 C     pepsi     1     1
#6 C     cola      1     1

这篇关于tidyverse pivot_longer 几组列，但避免中间 mutate_wider 步骤的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！