How to get this dcast'able long table in R? - r

Question

I am trying to apply dcast on long table, continua from the thread answer How to get this data structure in R?
Code
dat.m <- structure(c(150L, 60L, 41L, 61L, 0L, 0L), .Dim = c(3L, 2L), .Dimnames = list(
c("ave_max", "ave", "lepo"), NULL))
library("ggplot2")
library("data.table")
dat.m <- melt(as.data.table(dat.m, keep.rownames = "Vars"), id.vars = "Vars") # https://stackoverflow.com/a/44128640/54964
dat.m
print("New step")
# http://stackoverflow.com/a/44090815/54964
minmax <- dat.m[dat.m$Vars %in% c("ave_max","lepo"), ]
absol <- dat.m[dat.m$Vars %in% c("ave"), ]
#minm <- dcast(minmax, Vars ~ variable)
minm <- dcast(minmax, Vars ~ ...)
absol <- merge(absol, minm, by = "Vars", all.x = T)
absol
#Test function
ggplot(absol, aes(x = Vars, y = value, fill = variable)) +
geom_bar(stat = "identity") +
geom_errorbar(aes(ymin = lepo, ymax = ave_max), width = .25)
Output
dcast, melt
Vars variable value
1: ave_max V1 150
2: ave V1 60
3: lepo V1 41
4: ave_max V2 61
5: ave V2 0
6: lepo V2 0
[1] "New step"
Vars variable value V1 V2
1: ave V1 60 NA NA
2: ave V2 0 NA NA
Error in FUN(X[[i]], ...) : object 'lepo' not found
Calls: <Anonymous> ... by_layer -> f -> <Anonymous> -> f -> lapply -> FUN -> FUN
Execution halted
Expected output: to pass the test function ggplot
Testing Uwe's proposal
Aim is to get to this data structure
dat.m <- structure(c(150L, 60L, 41L, 61L, 0L, 0L), .Dim = c(3L, 2L), .Dimnames = list(c("ave_max", "ave", "lepo"), NULL))
from this data structure
dat.m <- structure(list(ave_max = c(15L, 6L), ave = c(6L, NA), lepo = c(4L, NA)), .Names = c("ave_max", "ave", "lepo"), class = "data.frame", row.names = c(NA, -2L))
Attempts
dat.m <- structure(list(ave_max = c(15L, 6L), ave = c(6L, NA), lepo = c(4L, NA)), .Names = c("ave_max", "ave", "lepo"), class = "data.frame", row.names = c(NA, -2L))
# ...
Code and output
dat.m <- setDT(dat.m)
Output wrong
ave_max ave lepo
1: 15 6 4
2: 6 NA NA
Classes ‘data.table’ and 'data.frame': 2 obs. of 3 variables:
$ ave_max: int 15 6
$ ave : int 6 NA
$ lepo : int 4 NA
- attr(*, ".internal.selfref")=<externalptr>
Code and output
dat.m <- as.matrix(dcast(melt(setDT(dat.m), measure.vars = names(dat.m)), variable ~ rowid(variable))[, variable := NULL]);
dimnames(dat.m) <- list(names(dat.m), NULL);
Output wrong
Error in `:=`(variable, NULL) :
Check that is.data.table(DT) == TRUE. Otherwise, := and `:=`(...) are defined for use in j, once only and in particular ways.
See help(":=").
R: 3.4.0 (backports)
OS: Debian 8.7.

Answer 1

The OP has edited his question and is supplying the data as a data.frame:
dat.df <- structure(list(ave_max = c(15L, 6L), ave = c(6L, NA), lepo = c(4L, NA)),
.Names = c("ave_max", "ave", "lepo"), class = "data.frame",
row.names = c(NA, -2L))
dat.df
# ave_max ave lepo
#1 15 6 4
#2 6 NA NA
class(dat.df)
#[1] "data.frame"
He is now asking to transform this data.frame into a matrix which is similar to the one used as input data in this answer.
This can be achieved by using data.table:
library(data.table) # CRAN version 1.10.4 used
# transpose the input data frame, use rowid() to create columns,
# remove a character column to ensure matrix will be of type integer,
# finally, coerce to matrix
dat.m2 <- as.matrix(
data.table::dcast(
data.table::melt(setDT(dat.df), measure.vars = names(dat.df)),
variable ~ rowid(variable)
)[, variable := NULL]
)
# add row names, remove column names
dimnames(dat.m2) <- list(names(dat.df), NULL)
dat.m2
# [,1] [,2]
#ave_max 15 6
#ave 6 NA
#lepo 4 NA
str(dat.m2)
# int [1:3, 1:2] 15 6 4 6 NA NA
# - attr(*, "dimnames")=List of 2
# ..$ : chr [1:3] "ave_max" "ave" "lepo"
# ..$ : NULL
class(dat.m2)
#[1] "matrix"
Edit: I've amended above code to use the double colon operator to explicitely state the namespace from which melt() and dcast() should be taken. Normally, this wouldn't be necessary as data.table is already loaded. However, the OP is reporting issues which might be caused by package reshape2 being loaded after data.table. The data.table package has it's own faster implementations of reshape2::dcast() and reshape2::melt(). When both packages have been loaded for some reason name clashes might occur.

Answer 2

The OP has supplied data as a matrix:
dat.m <- structure(c(150L, 60L, 41L, 61L, 0L, 0L), .Dim = c(3L, 2L), .Dimnames = list(
c("ave_max", "ave", "lepo"), NULL))
# dat.m
# [,1] [,2]
#ave_max 150 61
#ave 60 0
#lepo 41 0
class(dat.m)
#[1] "matrix"
For this data set, the OP wants to use ggplot2 to create a bar chart with error bars where the height of the bars is given by the values of ave and the lower and upper limits of the error bars by lepo and ave_max, resp., in each column.
As ggplot2 expects data to be supplied as data.frame the data needs to be transformed. For this, data.table is used:
library(data.table) # CRAN version 1.10.4 used
# convert to data.table & transpose
transposed <- dcast(melt(as.data.table(dat.m, keep.rownames = "Vars"),
id.vars = "Vars"), variable ~ ...)
setnames(transposed, "variable", "Vars")
library(ggplot2)
ggplot(transposed, aes(x = Vars, y = ave, ymin = lepo, ymax = ave_max)) +
geom_col() +
geom_errorbar(width = .25)