web-dev-qa-db-ja.com

複数の値列をワイドフォーマットに再形成する

次のデータフレームがあり、キャストを使用して2つの値(値とパーセント)の列を持つ「ピボットテーブル」を作成します。データフレームは次のとおりです。

expensesByMonth <- structure(list(month = c("2012-02-01", "2012-02-01", "2012-02-01", 
"2012-02-01", "2012-02-01", "2012-02-01", "2012-02-01", "2012-02-01", 
"2012-02-01", "2012-02-01", "2012-02-01", "2012-02-01", "2012-03-01", 
"2012-03-01", "2012-03-01", "2012-03-01", "2012-03-01", "2012-03-01", 
"2012-03-01", "2012-03-01", "2012-03-01", "2012-03-01", "2012-03-01", 
"2012-03-01", "2012-03-01", "2012-03-01", "2012-03-01", "2012-04-01", 
"2012-04-01", "2012-04-01", "2012-04-01", "2012-04-01", "2012-04-01", 
"2012-04-01", "2012-04-01", "2012-04-01", "2012-04-01", "2012-04-01", 
"2012-04-01", "2012-04-01", "2012-04-01", "2012-04-01", "2012-04-01", 
"2012-04-01", "2012-04-01", "2012-05-01", "2012-05-01", "2012-05-01", 
"2012-05-01", "2012-05-01", "2012-05-01", "2012-05-01", "2012-05-01", 
"2012-05-01", "2012-05-01", "2012-05-01", "2012-05-01", "2012-05-01", 
"2012-05-01", "2012-05-01", "2012-05-01", "2012-05-01", "2012-05-01", 
"2012-06-01", "2012-06-01", "2012-06-01", "2012-06-01", "2012-06-01", 
"2012-06-01", "2012-06-01", "2012-06-01", "2012-06-01", "2012-06-01", 
"2012-06-01", "2012-06-01", "2012-06-01", "2012-06-01", "2012-06-01", 
"2012-06-01", "2012-06-01", "2012-06-01", "2012-06-01", "2012-06-01", 
"2012-07-01", "2012-07-01", "2012-07-01", "2012-07-01", "2012-07-01", 
"2012-07-01", "2012-07-01", "2012-07-01", "2012-07-01", "2012-07-01", 
"2012-07-01", "2012-07-01", "2012-07-01"), 
expense_type = c("Adjustment", "Bank Service Charge", "Cable", "Clubbing", "Dining", "Education", 
"Gifts", "Groceries", "Lunch", "Personal Care", "Rent", "Transportation", 
"Adjustment", "Bank Service Charge", "Cable", "Clubbing", "Dining", 
"Gifts", "Groceries", "Lunch", "Medical Expenses", "Miscellaneous", 
"Personal Care", "Phone", "Recreation", "Rent", "Transportation", 
"Adjustment", "Bank Service Charge", "Clothes", "Clubbing", "Computer", 
"Dining", "Gifts", "Groceries", "Lunch", "Maintenance", "Medical Expenses", 
"Miscellaneous", "Personal Care", "Phone", "Recreation", "Rent", 
"Transportation", "Travel", "Bank Service Charge", "Cable", "Clothes", 
"Clubbing", "Computer", "Dining", "Electric", "Gifts", "Groceries", 
"Lunch", "Maintenance", "Medical Expenses", "Miscellaneous", 
"Personal Care", "Phone", "Recreation", "Rent", "Transportation", 
"Adjustment", "Bank Service Charge", "Cable", "Charity", "Clothes", 
"Computer", "Dining", "Education", "Electric", "Gifts", "Groceries", 
"Lunch", "Maintenance", "Medical Expenses", "Miscellaneous", 
"Personal Care", "Phone", "Recreation", "Rent", "Transportation", 
"Computer", "Gifts", "Groceries", "Lunch", "Maintenance", "Medical Expenses", 
"Miscellaneous", "Personal Care", "Phone", "Recreation", "Rent", 
"Repair and Maintenance", "Transportation"), 
value = c(442.37, 200, 21.33, 75, 22.5, 1800, 10, 233.33, 154.75, 30, 545, 32.5, 
2, 200, 36.33, 206.55, 74.5, 89, 372.68, 383.75, 144.19, 508.11, 
30, 38.4, 81.75, 1746.7, 35, 16.37, 200, 806.9, 324.81, 756, 
80.5, 100, 398.37, 326.25, 151, 29.95, 101, 90, 38.45, 61, 743.75, 
129, 228.53, 200, 39.05, 237, 40, 283.83, 141.32, 32.88, 30, 
424.4, 412, 142.75, 86.55, 1051.5, 30, 38.9, 51.5, 749.7, 35, 
10, 200, 16, 32.59, 149.81, 100, 80, 60, 31.91, 55, 397.25, 486.4, 
115.6, 47.08, 1000, 120, 41.11, 256, 761.6, 55, 10.54, 10, 342.11, 
291, 76.5, 66.8, 1008, 30, 41.11, 316, 765, 65, 62), 
percent = c(0.124025030980324, 0.0560729845967511, 0.00598018380724351, 0.0210273692237817, 
0.0063082107671345, 0.50465686137076, 0.00280364922983756, 0.0654175474797997, 
0.0433864718317362, 0.00841094768951267, 0.152798883026147, 0.00911185999697206, 
0.000506462461002391, 0.0506462461002391, 0.00919989060410842, 
0.0523049106600219, 0.018865726672339, 0.0225375795146064, 0.0943742149831854, 
0.0971774847048337, 0.0365134111259673, 0.128669320529962, 0.00759693691503586, 
0.0097240792512459, 0.0207016530934727, 0.442318990316438, 0.00886309306754183, 
0.00357276925628781, 0.0436502047194601, 0.176106750940662, 0.0708901149746392, 
0.164997773839559, 0.0175692073995827, 0.0218251023597301, 0.0869446602704567, 
0.0712043964486193, 0.0329559045631924, 0.00653661815673915, 
0.0220433533833274, 0.0196425921237571, 0.00839175185731621, 
0.0133133124394353, 0.162324198800492, 0.0281543820440518, 0.0498769064226911, 
0.0496724104530621, 0.00969853814096037, 0.0588618063868785, 
0.00993448209061241, 0.070492601294463, 0.0350985252261336, 0.0081661442784834, 
0.00745086156795931, 0.105404854981398, 0.102325165533308, 0.035453682960873, 
0.0214957356235626, 0.261152697956974, 0.00745086156795931, 0.00966128383312057, 
0.0127906456916635, 0.186197030583303, 0.00869267182928586, 0.00249044292527426, 
0.0498088585054852, 0.00398470868043882, 0.00811635349346881, 
0.0373093254635337, 0.0249044292527426, 0.0199235434021941, 0.0149426575516456, 
0.00794700337455016, 0.0136974360890084, 0.09893284520652, 0.12113514388534, 
0.0287895202161704, 0.0117250052921912, 0.249044292527426, 0.0298853151032911, 
0.0102382108658025, 0.0637553388870211, 0.189672133188888, 0.0136974360890084, 
0.00341757293956667, 0.0032424790697976, 0.110928451456846, 0.0943561409311103, 
0.0248049648839517, 0.021659760186248, 0.326841890235599, 0.00972743720939281, 
0.013329831455938, 0.102462338605604, 0.248049648839517, 0.0210761139536844, 
0.0201033702327451)), 
.Names = c("month", "expense_type", "value", "percent"), 
row.names = c(NA, -96L), 
class = "data.frame"
)

これは私が作成したいものです(もちろん、[月] _値、[月] _パーセントなどの異なるヘッダー名を使用):

expenses   value     percent value.1   percent.1 value.2   percent.2 value.3   percent.3 value.4   percent.4 value.5   percent.5
1              Adjustment  442.37 0.124025031    2.00 0.000506462   16.37 0.003572769    0.00 0.000000000   10.00 0.002490443    0.00 0.000000000
2     Bank Service Charge  200.00 0.056072985  200.00 0.050646246  200.00 0.043650205  200.00 0.049672410  200.00 0.049808859    0.00 0.000000000
3                   Cable   21.33 0.005980184   36.33 0.009199891    0.00 0.000000000   39.05 0.009698538   16.00 0.003984709    0.00 0.000000000
4                 Charity    0.00 0.000000000    0.00 0.000000000    0.00 0.000000000    0.00 0.000000000   32.59 0.008116353    0.00 0.000000000
5                 Clothes    0.00 0.000000000    0.00 0.000000000  806.90 0.176106751  237.00 0.058861806  149.81 0.037309325    0.00 0.000000000
6                Clubbing   75.00 0.021027369  206.55 0.052304911  324.81 0.070890115   40.00 0.009934482    0.00 0.000000000    0.00 0.000000000
7                Computer    0.00 0.000000000    0.00 0.000000000  756.00 0.164997774  283.83 0.070492601  100.00 0.024904429   10.54 0.003417573
8                  Dining   22.50 0.006308211   74.50 0.018865727   80.50 0.017569207  141.32 0.035098525   80.00 0.019923543    0.00 0.000000000
9               Education 1800.00 0.504656861    0.00 0.000000000    0.00 0.000000000    0.00 0.000000000   60.00 0.014942658    0.00 0.000000000
10               Electric    0.00 0.000000000    0.00 0.000000000    0.00 0.000000000   32.88 0.008166144   31.91 0.007947003    0.00 0.000000000
11                  Gifts   10.00 0.002803649   89.00 0.022537580  100.00 0.021825102   30.00 0.007450862   55.00 0.013697436   10.00 0.003242479
12              Groceries  233.33 0.065417547  372.68 0.094374215  398.37 0.086944660  424.40 0.105404855  397.25 0.098932845  342.11 0.110928451
13                  Lunch  154.75 0.043386472  383.75 0.097177485  326.25 0.071204396  412.00 0.102325166  486.40 0.121135144  291.00 0.094356141
14            Maintenance    0.00 0.000000000    0.00 0.000000000  151.00 0.032955905  142.75 0.035453683  115.60 0.028789520   76.50 0.024804965
15       Medical Expenses    0.00 0.000000000  144.19 0.036513411   29.95 0.006536618   86.55 0.021495736   47.08 0.011725005   66.80 0.021659760
16          Miscellaneous    0.00 0.000000000  508.11 0.128669321  101.00 0.022043353 1051.50 0.261152698 1000.00 0.249044293 1008.00 0.326841890
17          Personal Care   30.00 0.008410948   30.00 0.007596937   90.00 0.019642592   30.00 0.007450862  120.00 0.029885315   30.00 0.009727437
18                  Phone    0.00 0.000000000   38.40 0.009724079   38.45 0.008391752   38.90 0.009661284   41.11 0.010238211   41.11 0.013329831
19             Recreation    0.00 0.000000000   81.75 0.020701653   61.00 0.013313312   51.50 0.012790646  256.00 0.063755339  316.00 0.102462339
20                   Rent  545.00 0.152798883 1746.70 0.442318990  743.75 0.162324199  749.70 0.186197031  761.60 0.189672133  765.00 0.248049649
21 Repair and Maintenance    0.00 0.000000000    0.00 0.000000000    0.00 0.000000000    0.00 0.000000000    0.00 0.000000000   65.00 0.021076114
22         Transportation   32.50 0.009111860   35.00 0.008863093  129.00 0.028154382   35.00 0.008692672   55.00 0.013697436   62.00 0.020103370
23                 Travel    0.00 0.000000000    0.00 0.000000000  228.53 0.049876906    0.00 0.000000000    0.00 0.000000000    0.00 0.000000000

単一の値の列でキャストを使用しているときに次のエラーも発生しました。「値」パラメーターが考慮されていません。したがって、value = "percent"を指定しても、 "value"列の値が表示されます。

cast(expensesByMonth, expense_type ~ month, fun.aggregate = sum, value = "percent")
22
Alex Burdusel

最良のオプションは、meltを使用してデータを長い形式に再形成し、次にdcastに変換することです。

library(reshape2)

meltExpensesByMonth <- melt(expensesByMonth, id.vars=1:2)
dcast(meltExpensesByMonth, expense_type ~ month + variable, fun.aggregate = sum)

出力の最初の数行:

             expense_type 2012-02-01_value 2012-02-01_percent 2012-03-01_value 2012-03-01_percent
1              Adjustment           442.37        0.124025031             2.00       0.0005064625
2     Bank Service Charge           200.00        0.056072985           200.00       0.0506462461
3                   Cable            21.33        0.005980184            36.33       0.0091998906
4                 Charity             0.00        0.000000000             0.00       0.0000000000
25
Andrie

data.table は、複数のvalue.var変数にキャストできます。これは非常に直接的(かつ効率的)です。

したがって:

library(data.table) # v1.9.5+
dcast(setDT(expensesByMonth), expense_type ~ month, value.var = c("value", "percent"))
24
Arun

この質問はよく訪れるので、私の意見では完全なベースRの回答に値します。ベースRのreshape関数は非常に用途が広く、この問題にも簡単に適用できます。

expenses <- reshape(expensesByMonth, idvar = 'expense_type', direction = 'wide',
                    timevar = 'month', sep = '_')

NA- valuesのセルは、0で次のように置き換えることができます。

expenses[is.na(expenses)] <- 0

これは(目的の出力と比較しやすくするためにexpense_typeで並べ替えられます):

> expenses[order(expenses$expense_type),]
             expense_type value_2012-02-01 percent_2012-02-01 value_2012-03-01 percent_2012-03-01 value_2012-04-01 percent_2012-04-01 value_2012-05-01 percent_2012-05-01 value_2012-06-01 percent_2012-06-01 value_2012-07-01 percent_2012-07-01
1              Adjustment           442.37        0.124025031             2.00       0.0005064625            16.37        0.003572769             0.00        0.000000000            10.00        0.002490443             0.00        0.000000000
2     Bank Service Charge           200.00        0.056072985           200.00       0.0506462461           200.00        0.043650205           200.00        0.049672410           200.00        0.049808859             0.00        0.000000000
3                   Cable            21.33        0.005980184            36.33       0.0091998906             0.00        0.000000000            39.05        0.009698538            16.00        0.003984709             0.00        0.000000000
67                Charity             0.00        0.000000000             0.00       0.0000000000             0.00        0.000000000             0.00        0.000000000            32.59        0.008116353             0.00        0.000000000
30                Clothes             0.00        0.000000000             0.00       0.0000000000           806.90        0.176106751           237.00        0.058861806           149.81        0.037309325             0.00        0.000000000
4                Clubbing            75.00        0.021027369           206.55       0.0523049107           324.81        0.070890115            40.00        0.009934482             0.00        0.000000000             0.00        0.000000000
32               Computer             0.00        0.000000000             0.00       0.0000000000           756.00        0.164997774           283.83        0.070492601           100.00        0.024904429            10.54        0.003417573
5                  Dining            22.50        0.006308211            74.50       0.0188657267            80.50        0.017569207           141.32        0.035098525            80.00        0.019923543             0.00        0.000000000
6               Education          1800.00        0.504656861             0.00       0.0000000000             0.00        0.000000000             0.00        0.000000000            60.00        0.014942658             0.00        0.000000000
52               Electric             0.00        0.000000000             0.00       0.0000000000             0.00        0.000000000            32.88        0.008166144            31.91        0.007947003             0.00        0.000000000
7                   Gifts            10.00        0.002803649            89.00       0.0225375795           100.00        0.021825102            30.00        0.007450862            55.00        0.013697436            10.00        0.003242479
8               Groceries           233.33        0.065417547           372.68       0.0943742150           398.37        0.086944660           424.40        0.105404855           397.25        0.098932845           342.11        0.110928451
9                   Lunch           154.75        0.043386472           383.75       0.0971774847           326.25        0.071204396           412.00        0.102325166           486.40        0.121135144           291.00        0.094356141
37            Maintenance             0.00        0.000000000             0.00       0.0000000000           151.00        0.032955905           142.75        0.035453683           115.60        0.028789520            76.50        0.024804965
21       Medical Expenses             0.00        0.000000000           144.19       0.0365134111            29.95        0.006536618            86.55        0.021495736            47.08        0.011725005            66.80        0.021659760
22          Miscellaneous             0.00        0.000000000           508.11       0.1286693205           101.00        0.022043353          1051.50        0.261152698          1000.00        0.249044293          1008.00        0.326841890
10          Personal Care            30.00        0.008410948            30.00       0.0075969369            90.00        0.019642592            30.00        0.007450862           120.00        0.029885315            30.00        0.009727437
24                  Phone             0.00        0.000000000            38.40       0.0097240793            38.45        0.008391752            38.90        0.009661284            41.11        0.010238211            41.11        0.013329831
25             Recreation             0.00        0.000000000            81.75       0.0207016531            61.00        0.013313312            51.50        0.012790646           256.00        0.063755339           316.00        0.102462339
11                   Rent           545.00        0.152798883          1746.70       0.4423189903           743.75        0.162324199           749.70        0.186197031           761.60        0.189672133           765.00        0.248049649
95 Repair and Maintenance             0.00        0.000000000             0.00       0.0000000000             0.00        0.000000000             0.00        0.000000000             0.00        0.000000000            65.00        0.021076114
12         Transportation            32.50        0.009111860            35.00       0.0088630931           129.00        0.028154382            35.00        0.008692672            55.00        0.013697436            62.00        0.020103370
45                 Travel             0.00        0.000000000             0.00       0.0000000000           228.53        0.049876906             0.00        0.000000000             0.00        0.000000000             0.00        0.000000000

tidyverseを使用してこれを実現することもできます。

library(dplyr)
library(tidyr)

expensesByMonth %>% 
  gather(k, v, 3:4) %>% 
  unite(km, k, month) %>% 
  spread(km, v, fill = 0)
4
Jaap

私はこれのためにtabulateパッケージのtables関数を好みます。要素が必要ですが、これはとにかくあなたが持っているデータのタイプに関しては良い考えです。

library(tables)
expensesByMonth$month= as.factor(expensesByMonth$month)
expensesByMonth$expense_type= as.factor(expensesByMonth$expense_type)
tabular(expense_type~(month)*(value+percent)*(sum),data=expensesByMonth)
# Optional formatting
tabular(expense_type~month*
   ((Format(digits=1))*value+(Format(digits=3))*percent)*sum,
   data=expensesByMonth)

部分的な出力:

                       value      percent  value      percent  value      percent 
expense_type           sum        sum      sum        sum      sum        sum     
Adjustment              442       0.124025    2       0.000506   16       0.003573
Bank Service Charge     200       0.056073  200       0.050646  200       0.043650
Cable                    21       0.005980   36       0.009200    0       0.000000
3
Dieter Menne

Tidyr 1.0.0で導入された新しい関数pivot_wider()を使用して、複数の値/メジャー列を持つ長い形式からワイド形式への形状変更が可能になりました。

これは、以前のgather()のtidyr戦略よりもspread()よりも優れています。これは、属性が削除されなくなったためです(たとえば、日付は日付のままで、文字列は文字列のままです)。

pivot_wider()(対応するもの:pivot_longer())はspread()と同様に機能します。ただし、複数の値列などの追加機能を提供します。このため、引数values_from-どの列から値が取得されるかを示す-は、複数の列名をとることがあります。

NAsは、引数values_fillを使用して入力できます。

library("tidyr")
library("magrittr")

pivot_wider(expensesByMonth, 
            id_cols = expense_type,
            names_from = month,
            values_from = c(value, percent))
#> # A tibble: 23 x 13
#>    expense_type `value_2012-02-~ `value_2012-03-~ `value_2012-04-~
#>    <chr>                   <dbl>            <dbl>            <dbl>
#>  1 Adjustment              442.               2               16.4
#>  2 Bank Servic~            200              200              200  
#>  3 Cable                    21.3             36.3             NA  
#>  4 Clubbing                 75              207.             325. 
#>  5 Dining                   22.5             74.5             80.5
#>  6 Education              1800               NA               NA  
#>  7 Gifts                    10               89              100  
#>  8 Groceries               233.             373.             398. 
#>  9 Lunch                   155.             384.             326. 
#> 10 Personal Ca~             30               30               90  
#> # ... with 13 more rows, and 9 more variables: `value_2012-05-01` <dbl>,
#> #   `value_2012-06-01` <dbl>, `value_2012-07-01` <dbl>,
#> #   `percent_2012-02-01` <dbl>, `percent_2012-03-01` <dbl>,
#> #   `percent_2012-04-01` <dbl>, `percent_2012-05-01` <dbl>,
#> #   `percent_2012-06-01` <dbl>, `percent_2012-07-01` <dbl>

または、より細かい制御が可能なpivot specを使用して形状を変更することもできます(以下のリンクを参照)。

# see also ?build_wider_spec
spec <- expensesByMonth %>%
    expand(month, .value = c("percent", "value")) %>%
    dplyr::mutate(.name = paste(.$month, .$.value, sep = "_"))
pivot_wider_spec(expensesByMonth, spec = spec)

2019-03-26に reprexパッケージ (v0.2.1)によって作成されました

参照: https://tidyr.tidyverse.org/dev/articles/pivot.html

3
hplieninger