问题
I would like to transform a table like this (*):
set.seed(1)
mydata <- data.frame(ID=rep(1:4, each=3), R=rep(1:3, times=4), FIXED=rep(runif(4), each=3), AAA=rnorm(12), BBB=rbinom(12,12,0.5), CCC=runif(12))
ID R FIXED AAA BBB CCC
1 1 0.26 -0.83 8 0.82
1 2 0.26 1.59 5 0.64
1 3 0.26 0.32 6 0.78
2 1 0.37 -0.82 6 0.55
2 2 0.37 0.48 6 0.52
2 3 0.37 0.73 4 0.78
3 1 0.57 0.57 8 0.02
3 2 0.57 -0.30 7 0.47
3 3 0.57 1.51 7 0.73
4 1 0.90 0.38 4 0.69
4 2 0.90 -0.62 7 0.47
4 3 0.90 -2.21 6 0.86
Into wide format, like this:
ID FIXED AAA1 BBB2 CCC2 FIXED2 AAA2 BBB2 CCC2 FIXED3 AAA3 BBB3 CCC3
1 0.27 0.49 7 0.73 0.37 0.74 4 0.69 0.57 0.58 7 0.48
2 0.91 -0.31 6 0.86 0.20 1.51 8 0.44 0.90 0.39 7 0.24
3 0.94 -0.62 7 0.07 0.66 -2.21 6 0.10 0.63 1.12 6 0.32
4 0.06 -0.04 7 0.52 0.21 -0.02 3 0.66 0.18 0.94 6 0.41
How can I do it?
I've tried with
dcast(mydata, ID + FIXED ~ R, value.var=(names(mydata)[3:5])
or even writing the column names, "AAA", "BBB", "CCC", but it produces an error and I can't get the wide format I need. I've also tried other options, with no luck.
How can I do it?
(*) In reality has much more columns, but the story is the same.
The error is:
Error in .subset2(x, i, exact = exact) :
recursive indexing failed at level 2
In addition: Warning message:
In if (!(value.var %in% names(data))) { :
the condition has length > 1 and only the first element will be used
回答1:
You are referencing to the wrong value variables (the AAA, BBB and CCC columns have index numbers 4 - 6) and you should use setDT() to convert the dataframe to a datatable. Using:
dcast(setDT(mydata), ID + FIXED ~ R, value.var = names(mydata)[4:6])
which gives:
ID FIXED AAA_1 AAA_2 AAA_3 BBB_1 BBB_2 BBB_3 CCC_1 CCC_2 CCC_3
1: 1 0.2655087 -0.8356286 1.5952808 0.3295078 8 5 6 0.8209463 0.6470602 0.7829328
2: 2 0.3721239 -0.8204684 0.4874291 0.7383247 6 6 4 0.5530363 0.5297196 0.7893562
3: 3 0.5728534 0.5757814 -0.3053884 1.5117812 8 7 7 0.0233312 0.4772301 0.7323137
4: 4 0.9082078 0.3898432 -0.6212406 -2.2146999 4 7 6 0.6927316 0.4776196 0.8612095
If you don't convert to a datatable, the data.table package will fall back to the implementation of dcast from reshape2 which is not able to hande multiple value.var's, hence the error-message.
If you want another separator, you can add for example sep = '.' parameter to dcast.
回答2:
set.seed(1)
require(data.table)
mydata <- data.table(ID=rep(1:4, each=3), R=rep(1:3, times=4), FIXED=rep(runif(4), each=3), AAA=rnorm(12), BBB=rbinom(12,12,0.5), CCC=runif(12))
dcast(mydata, ID ~ R, value.var=names(mydata)[3:6])
ID FIXED_1 FIXED_2 FIXED_3 AAA_1 AAA_2 AAA_3 BBB_1 BBB_2 BBB_3 CCC_1 CCC_2 CCC_3
1: 1 0.43809711 0.43809711 0.43809711 -0.4781501 0.4179416 1.35867955 6 7 6 0.6422883 0.8762692 0.7789147
2: 2 0.24479728 0.24479728 0.24479728 -0.1027877 0.3876716 -0.05380504 5 7 5 0.7973088 0.4552745 0.4100841
3: 3 0.07067905 0.07067905 0.07067905 -1.3770596 -0.4149946 -0.39428995 7 4 5 0.8108702 0.6049333 0.6547239
4: 4 0.09946616 0.09946616 0.09946616 -0.0593134 1.1000254 0.76317575 4 5 3 0.3531973 0.2702601 0.9926841
来源：https://stackoverflow.com/questions/37622935/wide-format-with-dcast-data-table