Switch column to row in a data.frame

问题

I ran into an R problem, which seems to be a little bit tricky. I have a data.frame that looks like this:

Ident | A1 | ... | An | Z1 | ... | Zn
1     | 1  | ... | 1  | 1  | ... | 0
2     | 6  | ... | 4  | 0  | ... | 1
3     | 4  | ... | 4  | 1  | ... | 0
4     | 1  | ... | 4  | 0  | ... | 0

Now, what I want is to transform the original data.frame to the following structure:

Z     | A1 | ... | An
Z1    | 1  | ... | 1
Zn    | 6  | ... | 4
Z1    | 4  | ... | 4

Only rows are taken into the resulting data if any of the rows Z's is 1.

Any suggestions? A starting point may be sufficient. Many thanks in advance.

Well here is the dump:

structure(list(Ident = c(1, 2, 3, 4), A1 = c(1, 6, 4, 1), A2 = c(1, 
4, 4, 4), Z1 = c(1, 0, 1, 0), Z2 = c(0, 1, 0, 0)), .Names = c("Ident", 
"A1", "A2", "Z1", "Z2"), row.names = c(NA, -4L), class = "data.frame")

回答1:

You could write something like

dframe<-dframe[sum(dframe[,zindex1:zindexN])>0,Aindex1:AindexN]

where zindex1:zindexN is the range of column indices for Z, and similar for Aindex .

回答2:

Set up data:

edit: add an all-zero row.

dat <- structure(list(Ident = c(1, 2, 3, 4, 5), 
      A1 = c(1, 6, 4, 1, 2), A2 = c(1, 4, 4, 4, 3), 
      Z1 = c(1, 0, 1, 1, 0), Z2 = c(0, 1, 0, 0, 0)),
     .Names = c("Ident", "A1", "A2", "Z1", "Z2"), 
    row.names = c(NA, -5L), class = "data.frame")

Find out which columns have the Z elements:

Zcols <- grep("^Z[0-9]+",names(dat))

Pull out their names:

Znames <- names(dat)[Zcols]

Identify the relevant columns and get the appropriate names:

w <- apply(dat[Zcols],1,
           function(x) if (all(x==0)) NA else which(x==1))
dd <- data.frame(Z=Znames[w], dat[-Zcols])

If you like you can convert NA values:

levels(dd$Z) <- c(levels(dd$Z),"missing")
dd$Z[is.na(dd$Z)] <- "missing"

##         Z Ident A1 A2
## 1      Z1     1  1  1
## 2      Z2     2  6  4
## 3      Z1     3  4  4
## 4      Z1     4  1  4
## 5 missing     5  2  3

回答3:

Assuming that Ben's answer is what you're looking for (and using his sample data), perhaps you can use melt and merge, like this:

library(reshape2)
zCols <- grep("^Z", names(dat), value = TRUE)  ## Just the Z cols
otherCols <- setdiff(names(dat), zCols)        ## The other columns
datL <- melt(dat, measure.vars = zCols)        ## melting
merge(dat[otherCols],                          ## merging
      datL[as.logical(datL$value), c(otherCols, "variable")],
      all = TRUE)
#   Ident A1 A2 variable
# 1     1  1  1       Z1
# 2     2  6  4       Z2
# 3     3  4  4       Z1
# 4     4  1  4       Z1
# 5     5  2  3     <NA>

来源：https://stackoverflow.com/questions/16232192/switch-column-to-row-in-a-data-frame

标签

dataframe

transformation