Reputation: 103
I have a data frame with 144 columns and 1000 rows. It contains 36 different variables, always 4 values per variable - it looks that way:
1a-1d \t 2a-2d \t 3a-3d..........36a-36d
2 1 4 5 \t 3 4 5 3 \t 32 1 3 1.......3 12 4 1
.
.
4 5 2 6 \t 4 5 2 6 \t 23 5 2 5......3 1 5 6
What I want to do is to sum always a to d and name the output with the elements of a vector, e.g. names=c("AC_syn","AC_non",...)
.
I think the command must be somethin like:
ddply(a, .(), summarise, names[1]=a[,1]+a[,2]+a[,3]+a[,4], ...)
But there must be a more elegant way? Without too much copy-paste work? I am happy for every idea and little help!
Sorry that I did not make it clearer. Actually I wanted to have the sum of the columns after every fourth step (1-4, 5-8....141-144) and rename the new dataframe with the variables given in a vector. So for the input:
2 1 4 5 \t 3 4 5 3 \t 32 1 3 1.......3 12 4 1
It should reply
12 15 37.....20
Upvotes: 2
Views: 946
Reputation: 25914
If you want to sum every every four columns
# example data
set.seed(1)
(df <- data.frame(replicate(8,rnorm(5))))
X1 X2 X3 X4 X5 X6 X7
#1 -0.6264538 -0.8204684 1.5117812 -0.04493361 0.91897737 -0.05612874 1.35867955
#2 0.1836433 0.4874291 0.3898432 -0.01619026 0.78213630 -0.15579551 -0.10278773
#3 -0.8356286 0.7383247 -0.6212406 0.94383621 0.07456498 -1.47075238 0.38767161
#4 1.5952808 0.5757814 -2.2146999 0.82122120 -1.98935170 -0.47815006 -0.05380504
#5 0.3295078 -0.3053884 1.1249309 0.59390132 0.61982575 0.41794156 -1.37705956
X8
#1 -0.4149946
#2 -0.3942900
#3 -0.0593134
#4 1.1000254
#5 0.7631757
Create indicator for the columns to sum - this sums every four columns
(ind <- rep(1:2,each=4))
#[1] 1 1 1 1 2 2 2 2
Sum columns according to ind
t(rowsum(t(df),ind))
# 1 2
#[1,] 0.01992536 1.8065336
#[2,] 1.04472535 0.1292631
#[3,] 0.22529172 -1.0678292
#[4,] 0.77758346 -1.4212814
#[5,] 1.74295162 0.4238835
You can then use colnames
to assign column names.
Upvotes: 0
Reputation: 969
A fun way to do it relying on matrix multiplication:
First create an incidence matrix with only zeros and ones to post-multiply your data set (assuming it's called df):
M = matrix(0, 144, 36)
M = (row(M) >= {(col(M)-1)*4 + 1} & row(M) < {(col(M)-1)*4 + 5})*1
Then multiply M by df and name the columns:
sumvar = as.matrix(df) %*% M
names(sumvar) = c("AC_syn","AC_non",...)
sumvar will have 36 columns and 1000 rows. Just in case, M looks like this:
[,1] [,2] [,3] [,4]...
[1,] 1 0 0 0
[2,] 1 0 0 0
[3,] 1 0 0 0
[4,] 1 0 0 0
[5,] 0 1 0 0
[6,] 0 1 0 0
[7,] 0 1 0 0
[8,] 0 1 0 0
...
Upvotes: 1
Reputation: 835
Since your data is small, how about a FOR loop. This might be a crude way, but it works nevertheless -
j=seq(1,141,4)
k=j+3
for(i in 1:length(names))
{
new_col <- paste("sum_",i)
ds$new_col <- rowSums(ds[,j[i]:k[i]])
}
ds is your dataset name.
Upvotes: 0