R:使用两个列标题将Tid()收集到Tidy数据

时间:2018-10-08 09:55:42

标签: r dplyr

我想整理一些在前两行中用两个列标题设置的数据:

  • 第一行(标题):实际上是度量的类型(例如 估计值,标准错误,上限,下限)。

  • 第二行(也是标题):是度量的年份。

我是否可以使用gather()或其他任何方式整理数据?

此外,当重复测量时(例如Rank,Rank.1),它实际上应该仅读取Rank,并且仅因Year而不同。有什么办法可以解决这个问题?

  Country_Territory WBCode Estimate  StdErr NumSrc    Rank   Lower
1              Year   <NA>  1996.00 1996.00   1996 1996.00 1996.00
2           Andorra    ADO     1.32    0.48      1   87.10   72.04
3       Afghanistan    AFG    -1.29    0.34      2    4.30    0.00
4            Angola    AGO    -1.17    0.26      4    9.68    0.54
    Upper Estimate.1 StdErr.1 NumSrc.1  Rank.1 Lower.1 Upper.1
1 1996.00    1998.00  1998.00     1998 1998.00 1998.00 1998.00
2   96.77       1.38     0.46        1   89.18   74.74   96.91
3   27.42      -1.18     0.33        2    9.79    0.00   31.44
4   27.42      -1.41     0.21        6    1.55    0.00   13.40

这是输入我的数据样本的代码:

df <-  data.frame(stringsAsFactors=FALSE,
       Country_Territory = c("Year", "Andorra", "Afghanistan", "Angola"),
                  WBCode = c(NA, "ADO", "AFG", "AGO"),
                Estimate = c(1996, 1.32, -1.29, -1.17),
                  StdErr = c(1996, 0.48, 0.34, 0.26),
                  NumSrc = c(1996, 1, 2, 4),
                    Rank = c(1996, 87.1, 4.3, 9.68),
                   Lower = c(1996, 72.04, 0, 0.54),
                   Upper = c(1996, 96.77, 27.42, 27.42),
                Estimate = c(1998, 1.38, -1.18, -1.41),
                  StdErr = c(1998, 0.46, 0.33, 0.21),
                  NumSrc = c(1998, 1, 2, 6),
                    Rank = c(1998, 89.18, 9.79, 1.55),
                   Lower = c(1998, 74.74, 0, 0),
                   Upper = c(1998, 96.91, 31.44, 13.4)
    )

例如,此尝试:

df %>% gather(key = measure, value = number, 3:14)

没有完全满足我的需求

   Country_Territory WBCode    measure  number
1               Year   <NA>   Estimate 1996.00
2            Andorra    ADO   Estimate    1.32
3        Afghanistan    AFG   Estimate   -1.29
4             Angola    AGO   Estimate   -1.17
5               Year   <NA>     StdErr 1996.00
6            Andorra    ADO     StdErr    0.48

因为年份与Country_Territory混合在一起。

4 个答案:

答案 0 :(得分:2)

这是一个选择:

library(tidyverse)

# get unique Year values and create column names (to add later)
df %>%
  filter(Country_Territory == "Year") %>%
  gather() %>%
  filter(value != "Year" & !is.na(value)) %>%
  pull(value) %>%
  unique() %>%
  paste0("Year_",.) -> col_years

# reshape data (excluding the Year row)
df %>%
  filter(Country_Territory != "Year") %>%
  gather(key,y,-Country_Territory, -WBCode) %>%
  separate(key, c("measure","v")) %>%
  group_by(v = ifelse(is.na(v), 0, v)) %>%
  nest() -> df_info

reduce(df_info$data, function(x,y) left_join(x,y,by=c("Country_Territory","WBCode","measure"))) %>%
  setNames(c("Country_Territory", "WBCode", "measure", col_years))

# # A tibble: 18 x 5
#   Country_Territory WBCode measure  Year_1996 Year_1998
#   <chr>             <chr>  <chr>        <dbl>     <dbl>
# 1 Andorra           ADO    Estimate      1.32      1.38
# 2 Afghanistan       AFG    Estimate     -1.29     -1.18
# 3 Angola            AGO    Estimate     -1.17     -1.41
# 4 Andorra           ADO    StdErr        0.48      0.46
# 5 Afghanistan       AFG    StdErr        0.34      0.33
# 6 Angola            AGO    StdErr        0.26      0.21
# 7 Andorra           ADO    NumSrc        1         1   
# 8 Afghanistan       AFG    NumSrc        2         2   
# 9 Angola            AGO    NumSrc        4         6   
# 10 Andorra           ADO    Rank         87.1      89.2 
# 11 Afghanistan       AFG    Rank          4.3       9.79
# 12 Angola            AGO    Rank          9.68      1.55
# 13 Andorra           ADO    Lower        72.0      74.7 
# 14 Afghanistan       AFG    Lower         0         0   
# 15 Angola            AGO    Lower         0.54      0   
# 16 Andorra           ADO    Upper        96.8      96.9 
# 17 Afghanistan       AFG    Upper        27.4      31.4 
# 18 Angola            AGO    Upper        27.4      13.4 

如果将以上输出另存为df_upd,则可以重新整形以使Year作为一列:

df_upd %>%
  gather(Year, value, -Country_Territory, -WBCode, -measure) %>%
  separate(Year, c("y","Year"), convert = T) %>%
  select(-y)

# # A tibble: 36 x 5
#   Country_Territory WBCode measure   Year  value
#   <chr>             <chr>  <chr>    <int>  <dbl>
# 1 Andorra           ADO    Estimate  1996   1.32
# 2 Afghanistan       AFG    Estimate  1996  -1.29
# 3 Angola            AGO    Estimate  1996  -1.17
# 4 Andorra           ADO    StdErr    1996   0.48
# 5 Afghanistan       AFG    StdErr    1996   0.34
# 6 Angola            AGO    StdErr    1996   0.26
# 7 Andorra           ADO    NumSrc    1996   1   
# 8 Afghanistan       AFG    NumSrc    1996   2   
# 9 Angola            AGO    NumSrc    1996   4   
# 10 Andorra           ADO    Rank      1996  87.1 
# # ... with 26 more rows

答案 1 :(得分:1)

也许(如果您只有两种措施):

bind_rows(df %>% select(1:2,3:8) %>%
             mutate(Year=first(Estimate)),
          df %>% select(1:2,9:14) %>%
            rename_all(funs(str_replace(.,"\\.1",""))) %>%
            mutate(Year=first(Estimate))) %>%
  filter(Country_Territory!="Year")
#  Country_Territory WBCode Estimate StdErr NumSrc  Rank Lower Upper Year
#1           Andorra    ADO     1.32   0.48      1 87.10 72.04 96.77 1996
#2       Afghanistan    AFG    -1.29   0.34      2  4.30  0.00 27.42 1996
#3            Angola    AGO    -1.17   0.26      4  9.68  0.54 27.42 1996
#4           Andorra    ADO     1.38   0.46      1 89.18 74.74 96.91 1998
#5       Afghanistan    AFG    -1.18   0.33      2  9.79  0.00 31.44 1998
#6            Angola    AGO    -1.41   0.21      6  1.55  0.00 13.40 1998

答案 2 :(得分:1)

Data.table解决方案。

需要一些准备工作(设置名字并创建唯一名称表),但是速度很快。

解决方案也可以使用两年以上。

library( data.table )
dt <- as.data.table( df )  #or use setDT( df )

#extract unique years from the first row from the thirs column untill end of dt
dt.years <- as.data.table ( unique( t( (dt[1, 3:ncol(dt)]) ) ) )
dt.years[, year_id := 1:.N ]
setnames(dt.years, c("year", "year_id" ) )

#melt row 2:n of the data.table
dt.melt <- melt( dt[2:nrow(dt)], 
                 id.vars = c( "Country_Territory", "WBCode"),
                 measure = patterns( "Estimate", "StdErr", "NumSrc", "Rank", "Lower", "Upper"),
                 value.name = c( "Estimate", "StdErr", "NumSrc", "Rank", "Lower", "Upper" ),
                 variable.name = "year")

#left join both datatables
result <- dt.years[dt.melt, on = c( year_id = "year")]

#cleaning and renaming
result[, year_id := NULL]

结果

#    year Country_Territory WBCode Estimate StdErr NumSrc  Rank Lower Upper
# 1: 1996           Andorra    ADO     1.32   0.48      1 87.10 72.04 96.77
# 2: 1996       Afghanistan    AFG    -1.29   0.34      2  4.30  0.00 27.42
# 3: 1996            Angola    AGO    -1.17   0.26      4  9.68  0.54 27.42
# 4: 1998           Andorra    ADO     1.38   0.46      1 89.18 74.74 96.91
# 5: 1998       Afghanistan    AFG    -1.18   0.33      2  9.79  0.00 31.44
# 6: 1998            Angola    AGO    -1.41   0.21      6  1.55  0.00 13.40

答案 3 :(得分:1)

data.table的{​​{1}}方法能够同时重塑多个度量列。无需使用melt()函数来重命名列。

patterns()
library(data.table)
# reshape multiple measure columns simultaneously from wide to long format
cols <- c("Estimate", "StdErr", "NumSrc", "Rank", "Lower", "Upper")
long <- melt(setDT(df), measure.vars = patterns(cols), value.name = cols)
# extract years
yrs <- long[Country_Territory == "Year", .(variable, Year = as.integer(Estimate))]
# join to get a separate Year column, remove Year rows and helper column 
result <- yrs[long[Country_Territory != "Year"], on = "variable"][, variable := NULL][]
result

整形后, Year Country_Territory WBCode Estimate StdErr NumSrc Rank Lower Upper 1: 1996 Andorra ADO 1.32 0.48 1 87.10 72.04 96.77 2: 1996 Afghanistan AFG -1.29 0.34 2 4.30 0.00 27.42 3: 1996 Angola AGO -1.17 0.26 4 9.68 0.54 27.42 4: 1998 Andorra ADO 1.38 0.46 1 89.18 74.74 96.91 5: 1998 Afghanistan AFG -1.18 0.33 2 9.79 0.00 31.44 6: 1998 Angola AGO -1.41 0.21 6 1.55 0.00 13.40 列表示以宽格式属于一列子集的行,即属于某一特定年份的行。