如何根据R中的多个条件从大数据帧中提取不同长度的向量

Question

I have a data frame in R that consists of 3 columns. It looks a bit like this:

  x      id trialNumber
1 1.4788 subj_01    trial010
2 1.4794 subj_01    trial010
3 1.4823 subj_01    trial010
4 1.4845 subj_01    trial010
5 1.4889 subj_01    trial010
6 1.4901 subj_01    trial010
...
20121 -1.3597 subj_03    trial042
20122 -1.3601 subj_03    trial042
20123 -1.3667 subj_03    trial042
20124 -1.3713 subj_03    trial042
20125 -1.3800 subj_03    trial042
20126 -1.3857 subj_03    trial042

I want to create a new data frame that consists of multiple columns for x; where the columns are defined by id and trialNumber. The number of rows of each combination of id and trialNumber varies. The number of rows in the new data frame should correspond to the largest number of rows of all the id and trialNumber combinations. The result should look sth like this:

x1      x2   ... xi
1.4788  1.5678  ...
1.4794  1.5789  ...
1.4823  1.5984  ...
1.4845  ...     ...
1.4889  NA      ...
1.4901  NA      -1.3713
...     ...     -1.3800
NA      ...     -1.3857

x1 to xi in the new data frame should correspond to each unique combination of id and trialNumber in the original data frame, eg x1 would correspond to all x where id == 'subj01' and trialNumber == 'trial010'.

There are a lot of combinations of id and trialNumber, so I don't want to manually define the conditions by which to subset the original data frame.

Answer 1

You could try (a suggestion after reading the above comments):

tapply(df$x, paste0(df$id,df$trialNumber), function(x) data.frame(mean = mean(x), lower_limit = mean(x) - sd(x), upper_limit = mean(x) + sd(x)))
$subj_01trial010
      mean lower_limit upper_limit
1 1.484871    1.479965    1.489778

$subj_03trial042
       mean lower_limit upper_limit
1 -1.370583   -1.381177    -1.35999

Or using aggregate you get a nicer outpur format:

aggregate(x ~ id + trialNumber, data = df, FUN = function(x) c(mean = mean(x), lower_limit = mean(x) - sd(x), upper_limit = mean(x) + sd(x)))
       id trialNumber    x.mean x.lower_limit x.upper_limit
1 subj_01    trial010  1.484871      1.479965      1.489778
2 subj_03    trial042 -1.370583     -1.381177     -1.359990

Answer 2

Here's an approach if you really want columns of x for each combination of trial and subject bound together:

#step 1: create vector of x per combination

step1 <- split(dat2$x, list(dat2$trial,dat2$subject))

#calculate max length(to add padding)
max_length <- max(sapply(step1,length))

#make all vectors same length padded with NA
step2 <- lapply(step1, function(x){
  length(x) <- max_length
  x
})

#combine

res <- do.call(cbind,step2)
res

Code used for data generating:

set.seed(100)

dat1 <-expand.grid(trial=sprintf("trial_%.03d",1:10), 
                   subject= sprintf("subj_%.02d",1:3))

dat2 <- dat1[sample(nrow(dat1),1000,T),]
dat2$x <- rnorm(nrow(dat2))