Creating multiple columns by different functions of different variables

Question

I'm having trouble transitioning to data.table. I am trying to group by some categorical variables, and apply a list of functions that each target different variables in order to create new columns. This is something that seems like it should be easy with mapply or Map, but I can't figure out to assemble the proper subset in the to pass to the functions.

Here is what it looks like,

set.seed(2015)
dat <- data.table(cat1 = factor('Total'),
                  cat2 = factor(rep(letters[1:4], 5)),
                  cat3 = factor(rep(1:4, each=5)),
                  var1 = sample(20),
                  var2 = sample(20),
                  var3 = sample(20))

## I have list of factor columns to group by
groups <- c(paste0("cat", 1:3))
setkeyv(dat, groups)

## List of functions, and corresponding list of column names that
## they are to be applied to.  So, in this example I should get
## two new columns: V1=sum(var1) and V2=mean(var2, var3)
thing <- function(...) mean(c(...), na.rm=TRUE)  # arbitrary function
funs <- list("sum", "thing")                     # named functions
targets <- list("var1", c("var2", "var3"))       # variables
outnames <- funs                                 # names or result columns

## Can't get this part
f <- function(fn, vars) do.call(fn, vars)
dat[, outnames := Map(f, funs, targets), by=groups]

The result for this example should be like this

dat[, `:=`(sum=sum(var1), thing=thing(var2, var3)), by=groups]

MichaelChirico · Accepted Answer · 2020-04-24 06:40:54Z

3

We need to subset the dataset columns based on the column names in the 'targets' list. One way would to loop through the list elements of 'targets' and subset the data.table (.SD[, x, with=FALSE]), and then apply the function.

dat[, unlist(outnames) := Map(f, funs, lapply(targets, function(x)
                          .SD[, ..x])), by = groups]

edited Apr 24, 2020 at 6:40

MichaelChirico

34.9k17 gold badges122 silver badges209 bronze badges

answered Nov 8, 2015 at 17:02

akrun

891k38 gold badges590 silver badges700 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Rorschach Over a year ago

great, looks good, I might have a follow up question in a bit, since I also need to be subsetting by indices at the same time. Thanks!

Collectives™ on Stack Overflow

Creating multiple columns by different functions of different variables

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related