如何将宽范围的连续数据重整为长类别的数据？

Question

我的数据采用以下宽格式，根据SUBJECT_ID按行显示，总共观察到变量X和Y ，然后观察到元数据的各个列，例如SUBJECT_BIRTHYEAR ， SUBJECT_HOMETOWN ：

variableX    variableY    SUBJECT_ID     SUBJECT_BIRTHYEAR     SUBJECT_HOMETOWN
2            1            A              1950                  Townsville
1            2            B              1951                  Villestown

我想将它们转换为以下长格式，其中对于每个SUBJECT_ID变量X和Y每次观察：

VARIABLE     SUBJECT_ID     SUBJECT_BIRTHYEAR     SUBJECT_HOMETOWN
X            A              1950                  Townsville
X            A              1950                  Townsville
Y            A              1950                  Townsville
X            B              1951                  Villestown
Y            B              1951                  Villestown
Y            B              1951                  Villestown

我的问题的具体内容是如何将n个连续变量的观测值转换为n行分类数据。

Answer 1

尝试以下

数据

df <- read.table(text="variableX    variableY    SUBJECT_ID     SUBJECT_BIRTHYEAR     SUBJECT_HOMETOWN
2            1            A              1950                  Townsville
1            2            B              1951                  Villestown", header=TRUE)

解

library(tidyverse)
result <- df %>%
        nest(variableX, variableY, .key="VARIABLE") %>%
        mutate(VARIABLE = map(VARIABLE, function(i) {
                                    vec <- unlist(i)
                                    rep(gsub("variable", "", names(vec)), times=vec)
                                })) %>%
        unnest()

# A tibble: 6 x 4
  # SUBJECT_ID SUBJECT_BIRTHYEAR SUBJECT_HOMETOWN VARIABLE
      # <fctr>             <int>           <fctr>    <chr>
# 1          A              1950       Townsville        X
# 2          A              1950       Townsville        X
# 3          A              1950       Townsville        Y
# 4          B              1951       Villestown        X
# 5          B              1951       Villestown        Y
# 6          B              1951       Villestown        Y

Answer 2

该问题要求dcast()的调用反转，该调用已使用length()作为聚合函数将数据从长格式整形为宽格式。

这可以通过调用melt()以及一些其他转换来实现：

library(data.table)
# reshape wide back to long format
long <- melt(setDT(wide), measure.vars = c("variableX", "variableY"))[
  # undo munging of variable names
  , variable := stringr::str_replace(variable, "^variable", "")][]
# undo effect of aggregation by length()
result <- long[long[, rep(.I, value)]][
  # beautify result
  order(SUBJECT_ID), !"value"]
result

  SUBJECT_ID SUBJECT_BIRTHYEAR SUBJECT_HOMETOWN variable 1: A 1950 Townsville X 2: A 1950 Townsville X 3: A 1950 Townsville Y 4: B 1951 Villestown X 5: B 1951 Villestown Y 6: B 1951 Villestown Y 

.I是保存行位置（即行索引）的特殊符号。

---

为了证明这确实是逆运算，可以重新调整result wide以重现：

dcast(result, ... ~ paste0("variable", variable), length, value.var = "variable")

  SUBJECT_ID SUBJECT_BIRTHYEAR SUBJECT_HOMETOWN variableX variableY 1: A 1950 Townsville 2 1 2: B 1951 Villestown 1 2 

数据

library(data.table)
wide <- fread("variableX    variableY    SUBJECT_ID     SUBJECT_BIRTHYEAR     SUBJECT_HOMETOWN
2            1            A              1950                  Townsville
1            2            B              1951                  Villestown")

Answer 3

这是使用base R的选项

res <- cbind(VARIABLE = rep(substr(names(df1)[1:2], 9, 9)[row(df1[1:2])], t(df1[1:2])), 
        df1[rep(seq_len(nrow(df1)), rowSums(df1[1:2])), -(1:2)])
row.names(res) <- NULL
res
#   VARIABLE SUBJECT_ID SUBJECT_BIRTHYEAR SUBJECT_HOMETOWN
#1        X          A              1950       Townsville
#2        X          A              1950       Townsville
#3        Y          A              1950       Townsville
#4        X          B              1951       Villestown
#5        Y          B              1951       Villestown
#6        Y          B              1951       Villestown