Condense/merge cells in a table in R - r

Question

I'm trying to do something with a table in R.
The table comes into the script like this
M P
Position1 34 56
Position2 45 23
Position3 89 78
Position1 56 45
Position3 54 35
Position2 56 89
And after analyzing this script, ideally, I'd like a final output to be this:
M P
Position1 90 101
Position2 101 102
Position3 143 113
Basically I sum the total number across the positions for M and P. I was wondering if there was an easier way to do this. The positions will be at random. Is there a way to potentially split the data table by the position?

Answer 1

You can use summarise_each from dplyr if you have multiple columns and you have a big dataset and of course the data is data.frame (From the post, it is not clear whether you have a matrix or data.frame)
library(dplyr)
dat %>%
group_by(Pos) %>%
summarise_each(funs(sum=sum(., na.rm=TRUE)))
# Pos M P
#1 Position1 90 101
#2 Position2 101 112
#3 Position3 143 113
Or another option I would use for bigger datasets is data.table. From the benchmarks by #Ananda Mahto, it is the clear winner in speed.
library(data.table)
setDT(dat)[, lapply(.SD, sum, na.rm=TRUE), by=Pos]
# Pos M P
#1: Position1 90 101
#2: Position2 101 112
#3: Position3 143 113
If you are using a matrix and do not need to transform it to data.frame with creating a new column for row.names. (Perhaps, that option would still be efficient)
do.call(rbind, by(m1, list(rownames(m1)), colSums, na.rm=TRUE))
# M P
#Position1 90 101
#Position2 101 112
#Position3 143 113
Or a slightly more efficient method when dealing with matrices
library(reshape2)
acast(melt(m1), Var1~Var2, value.var="value", sum, na.rm=TRUE)
# M P
#Position1 90 101
#Position2 101 112
#Position3 143 113
data
The rownames are added as a column as data.frame won't allow duplicate rownames.
dat <- structure(list(Pos = c("Position1", "Position2", "Position3",
"Position1", "Position3", "Position2"), M = c(34L, 45L, 89L,
56L, 54L, 56L), P = c(56L, 23L, 78L, 45L, 35L, 89L)), .Names = c("Pos",
"M", "P"), class = "data.frame", row.names = c(NA, -6L))
m1 <- structure(c(34, 45, 89, 56, 54, 56, 56, 23, 78, 45, 35, 89), .Dim = c(6L,
2L), .Dimnames = list(c("Position1", "Position2", "Position3",
"Position1", "Position3", "Position2"), c("M", "P")))

Answer 2

One more, just for fun. This one produces the structure you show in the post.
t(sapply(split(dat[-1], dat$Pos), colSums))
# M P
# Position1 90 101
# Position2 101 112
# Position3 143 113

Answer 3

This answer only applies if you are dealing with a matrix (like the "m1" dataset shared in #akrun's answer):
xtabs(Freq ~ Var1 + Var2, data.frame(as.table(m1)))
# Var2
# Var1 M P
# Position1 90 101
# Position2 101 112
# Position3 143 113

Answer 4

'aggregate' which needs to be used as follows:
> ddf
V1 V2 V3
1 Position1 34 56
2 Position2 45 23
3 Position3 89 78
4 Position1 56 45
5 Position3 54 35
6 Position2 56 89
> a1 = aggregate(V2~V1, ddf, sum)
> a2 = aggregate(V3~V1, ddf, sum)
> merge(a1, a2)
V1 V2 V3
1 Position1 90 101
2 Position2 101 112
3 Position3 143 113

Answer 5

First get your rownames
rows<-unique(rownames(yourDataFrame))
Make sure unique is there or we'll get a lot of duplicates
Then you can do a couple of different things here, the package plyr would come in handy, but just using base R you can use lapply to calculate your sums
result<-lapply(rownames, function(rname){
subsetDF<-yourDataFrame[rname,]
apply(subsetDF, 2, sum)
}
)
To break it down, you take all your rownames, and in lapply subset your dataframe by just the rows of that rowname. Next, you apply sum over that subset, taking the columns, and then output that to a list. You could then do something like rbindlist(result) to get your resulting dataframe.
Definitely not the most efficient way to do it, but it's the first thing I thought of

Answer 6

What you want is the aggregate function.
Say you have your table stored as data then try
condensedData <- aggregate(data, by=list(position), FUN=sum, na.rm=TRUE)
If that doesn't do exactly what you want, try experimenting with the aggregate function. The important inputs are by and FUN. by tells aggregate which columns you want the result to be identified uniquely by, while FUN tells aggregate what to do to combine numbers with the same by. FUN can be "sum", "mean", etc...