从宽到长的重塑，每个类别的子类别数量可变

Question

我有一个具有此类内容的数据集：

data <- read.table(header=TRUE, text='ID    Age Gender  M1.Date M1.Code M2.Date M2.Code M3.Other    M3.Code M3.Data
                   ID1  34  Male    23-Oct-18   M1  02-Apr-18   M2  31  M3  14-Jul-18
                   ID1  34  Male    02-Sep-19   M1  12-May-18   M2  23  M3  03-Nov-18
                   ID1  34  Male    NA  NA  03-Dec-18   M2  85  M3  04-Oct-19
                   ID1  34  Male    NA  NA  31-May-19   M2  NA  NA  NA
                   ID2  21  Female  13-May-18   M1  24-Jun-18   M2  734 M3  31-Aug-18
                   ID2  21  Female  21-Dec-18   M1  NA  NA  12  M3  08-Apr-19
                   ID2  21  Female  NA  NA  NA  NA  14  M3  16-Aug-19')

对于每个个体（ID），都有关于三种表型（Data，Code和Other）的信息（M1，M2，M3）。但是，每个表型的信息数量和类型因人而异。

为了分析数据（数百个个体），我想将数据集从宽转换为长。

理想情况下，输出看起来像这样：

output <- read.table(header=TRUE, text='ID  Age Gender  Code    Date    Other
                     ID1    34  Male    F1  23-Oct-2018 NA
                     ID1    34  Male    F1  02-Sept-2019    NA
                     ID1    34  Male    F2  02-Apr-2018 NA
                     ID1    34  Male    F2  12-May-2018 NA
                     ID1    34  Male    F2  03-Dec-2018 NA
                     ID1    34  Male    F2  31-May-2019 NA
                     ID1    34  Male    F3  14-Jul-18   31
                     ID1    34  Male    F3  03-Nov-18   23
                     ID1    34  Male    F3  04-Oct-19   85
                     ID2    21  Female  F1  13-May-2018 NA
                     ID2    21  Female  F1  21-Dec-2018 NA
                     ID2    21  Female  F2  24-Jun-2018 NA
                     ID2    21  Female  F3  31-Aug-18   734
                     ID2    21  Female  F3  08-Apr-19   12
                     ID2    21  Female  F3  16-Aug-19   14')

我应该如何进行？我尝试在R中使用重塑函数，但是由于每种表型的信息数量各不相同，因此我无法使其正常工作。

Answer 1

这也可以使用pivot_longer通过tidyr完成。

由于不同变量缺少一些列（例如，您没有M2.Other或M3.Other，只有M1.Other），因此您需要为这些变量添加额外的列并填充NA 。

如果变量以数字结尾而不是中间（将变量命名为MDate_1而不是M1.Date，这也将变得更加容易）。但是可以将变量名重命名为这种效果。

我希望这会有所帮助。

library(tidyr)

data <- read.table(header=TRUE, text='ID    Age Gender  M1.Date M1.Code M2.Date M2.Code M3.Other    M3.Code M3.Date
                   ID1  34  Male    23-Oct-18   M1  02-Apr-18   M2  31  M3  14-Jul-18
                   ID1  34  Male    02-Sep-19   M1  12-May-18   M2  23  M3  03-Nov-18
                   ID1  34  Male    NA  NA  03-Dec-18   M2  85  M3  04-Oct-19
                   ID1  34  Male    NA  NA  31-May-19   M2  NA  NA  NA
                   ID2  21  Female  13-May-18   M1  24-Jun-18   M2  734 M3  31-Aug-18
                   ID2  21  Female  21-Dec-18   M1  NA  NA  12  M3  08-Apr-19
                   ID2  21  Female  NA  NA  NA  NA  14  M3  16-Aug-19')

# Recode to have numbers at end after underscore instead of embedded in middle
names(data) <- sub("M(\\d).(\\w*)", "M\\2_\\1", names(data))

# Get all combinations of variables since there are missing columns
all_vars <- expand_grid(
  value = c("MDate", "MCode", "MOther"),
  time = 1:3
) %>% unite("vars", everything())

# Get missing vars and set to NA so have complete variables for pivot_longer
missing_vars <- setdiff(all_vars$vars, names(data))
data[missing_vars] <- list(NA)

# Convert variables to data types and make data longer
data %>%
  mutate_at(vars(starts_with("MDate")), funs(as.Date(., "%d-%b-%y")),
            vars(starts_with("MCode")), funs(as.character),
            vars(starts_with("MOther")), funs(as.character)) %>%
  pivot_longer( 
             cols = -c(ID, Age, Gender), 
             names_to = c(".value", "time"),
             names_sep = "_",
             values_drop_na = TRUE)

# A tibble: 15 x 7
   ID      Age Gender time  MDate      MCode MOther
   <fct> <int> <fct>  <chr> <date>     <fct>  <int>
 1 ID1      34 Male   1     2018-10-23 M1        NA
 2 ID1      34 Male   2     2018-04-02 M2        NA
 3 ID1      34 Male   3     2018-07-14 M3        31
 4 ID1      34 Male   1     2019-09-02 M1        NA
 5 ID1      34 Male   2     2018-05-12 M2        NA
 6 ID1      34 Male   3     2018-11-03 M3        23
 7 ID1      34 Male   2     2018-12-03 M2        NA
 8 ID1      34 Male   3     2019-10-04 M3        85
 9 ID1      34 Male   2     2019-05-31 M2        NA
10 ID2      21 Female 1     2018-05-13 M1        NA
11 ID2      21 Female 2     2018-06-24 M2        NA
12 ID2      21 Female 3     2018-08-31 M3       734
13 ID2      21 Female 1     2018-12-21 M1        NA
14 ID2      21 Female 3     2019-04-08 M3        12
15 ID2      21 Female 3     2019-08-16 M3        14