我需要将一个宽数据集转换为长数据集，并且有16列必须收敛到4列。每4列包含彼此相关的信息，并且该信息在转换中一定不能“丢失”。
我有来自四个区块的排名任务的数据，该任务实质上给了我一个数据集，其中信息以较宽的格式分为四个组。即first_image，first_sex，first_score，second_image，second_sex，second_score ...
我尝试了group_by和collect（）的各种组合，但距离还很遥远。
我已经读过Reshaping multiple sets of measurement columns (wide format) into single columns (long format)，但我恐怕没有一个聪明的人。
我已经制作了一些参与者数据的样例数据，并且还样例了我希望数据如何显示。
library(tidyverse)
sample_dat <- data.frame(subject_id = rep("sj1", 4),
first_pick = rep(1, 4),
first_image_pick = (c("a", "b", "c", "d")),
first_pick_neuro = rep("TD", 4),
first_pick_sex = rep("F", 4),
second_pick = rep(2, 4),
second_image_pick = (c("e", "f", "g", "h")),
second_pick_neuro = rep("TD", 4),
second_pick_sex = rep("M", 4),
third_pick = rep(3, 4),
third_image_pick = (c("i", "j", "k", "l")),
third_pick_neuro = rep("DS", 4),
third_pick_sex = rep("F", 4),
fourth_pick = rep(4, 4),
fourth_image_pick = (c("m", "n", "o", "p")),
fourth_pick_neuro = rep("DS", 4),
fourth_pick_sex = rep("M", 4))
预期输出：
final_data <- data.frame(subject_id = rep("sj1", 16),
image = c("a", "b", "c", "d",
"e", "f", "g", "h",
"i", "j", "k", "l",
"m", "n", "o", "p"),
rank = rep(c(1, 2, 3, 4), each = 4), # from the numbers in the first_pick, second_pick etc.
neuro = rep(c("TD", "DS"), each = 8),
sex = rep(c("F", "M", "F", "M"), each = 4))
到目前为止，我已经尝试过了，但是它只复制了所有信息：
sample_dat_long <- sample_dat %>%
group_by(subject_id) %>%
gather(Pick, Image,
first_image_pick,
second_image_pick,
third_image_pick,
fourth_image_pick)
因此，从本质上讲，我不想在收集数据时丢失每个图像（拾取，性别，神经）的信息。
任何帮助都会很棒！
答案 0 :(得分：2)
我们可以使用melt中的data.table来完成此操作，这可能需要多个measure patterns从“宽”格式改成“长”格式。在这里，带有子字符串'image'，'neuro'，'sex'的列名称被重塑为单独的列，以获得预期的输出
library(data.table)
melt(setDT(sample_dat), measure = patterns("image", "neuro", "sex"),
value.name = c("image", "neuro", "sex"), variable.name = 'rank')[,
.(subject_id, rank, image, neuro, sex)]
答案 1 :(得分：1)
我想您可以逐列进行操作，因为最后只需要4列。 获取应该放在第一个列中的列的索引（如果我理解正确的话）：
ind1 = seq(2,length(sample_dat[1,]), 4)
column1 = gather( sample_dat[,ind1] )[2]
然后对所有其他3列重复：
ind2 = seq(3,length(sample_dat[1,]), 4)
column2 = gather( sample_dat[,ind2] )[2]
您甚至可以使用for循环来完成这4列，而不必“手动”执行。 然后将它们组合回数据框
答案 2 :(得分：1)
值得考虑使用合适的列名（即"<variable_chr>.<time_num>"）。但是我们可以在一秒钟内解决它。
pfx <- c("first", "second", "third", "fourth")
names(sample_dat)[-1] <- sapply(names(sample_dat)[-1], function(x) {
x <- gsub("_pick", "", x)
if (lengths(strsplit(x, "_")) == 2)
sub("(^.*)_(.*)", paste("\\2", which(pfx == sub("(^.*)_.+", "\\1", x)), sep="."), x)
else
paste0("rank.", which(pfx == x))
})
names(sample_dat) # good names now
# [1] "subject_id" "rank.1" "image.1" "neuro.1" "sex.1" "rank.2"
# [7] "image.2" "neuro.2" "sex.2" "rank.3" "image.3" "neuro.3"
# [13] "sex.3" "rank.4" "image.4" "neuro.4" "sex.4"
此后，我们可以轻松使用reshape。
reshape(sample_dat, idvar="subject_id", varying=2:17, direction="long",
new.row.names=seq(ncol(sample_dat) - 1))
# subject_id time rank image neuro sex
# 1 sj1 1 1 a TD F
# 2 sj1 1 1 b TD F
# 3 sj1 1 1 c TD F
# 4 sj1 1 1 d TD F
# 5 sj1 2 2 e TD M
# 6 sj1 2 2 f TD M
# 7 sj1 2 2 g TD M
# 8 sj1 2 2 h TD M
# 9 sj1 3 3 i DS F
# 10 sj1 3 3 j DS F
# 11 sj1 3 3 k DS F
# 12 sj1 3 3 l DS F
# 13 sj1 4 4 m DS M
# 14 sj1 4 4 n DS M
# 15 sj1 4 4 o DS M
# 16 sj1 4 4 p DS M
sample_dat <- structure(list(subject_id = structure(c(1L, 1L, 1L, 1L), .Label = "sj1", class = "factor"),
first_pick = c(1, 1, 1, 1), first_image_pick = structure(1:4, .Label = c("a",
"b", "c", "d"), class = "factor"), first_pick_neuro = structure(c(1L,
1L, 1L, 1L), .Label = "TD", class = "factor"), first_pick_sex = structure(c(1L,
1L, 1L, 1L), .Label = "F", class = "factor"), second_pick = c(2,
2, 2, 2), second_image_pick = structure(1:4, .Label = c("e",
"f", "g", "h"), class = "factor"), second_pick_neuro = structure(c(1L,
1L, 1L, 1L), .Label = "TD", class = "factor"), second_pick_sex = structure(c(1L,
1L, 1L, 1L), .Label = "M", class = "factor"), third_pick = c(3,
3, 3, 3), third_image_pick = structure(1:4, .Label = c("i",
"j", "k", "l"), class = "factor"), third_pick_neuro = structure(c(1L,
1L, 1L, 1L), .Label = "DS", class = "factor"), third_pick_sex = structure(c(1L,
1L, 1L, 1L), .Label = "F", class = "factor"), fourth_pick = c(4,
4, 4, 4), fourth_image_pick = structure(1:4, .Label = c("m",
"n", "o", "p"), class = "factor"), fourth_pick_neuro = structure(c(1L,
1L, 1L, 1L), .Label = "DS", class = "factor"), fourth_pick_sex = structure(c(1L,
1L, 1L, 1L), .Label = "M", class = "factor")), class = "data.frame", row.names = c(NA,
-4L))