Efficient Collapse Dummy Variables

Question

What is an efficient way (any solution including non-base packages welcomed) to collapse dummy variables back into a factor.

   race.White race.Hispanic race         


        

Answer 1

We can use max.col to get the column index, subset the column names based on that and use sub to remove the prefix.

sub('[^.]+\\.', '', names(dat)[max.col(dat)])
#[1] "White"    "Asian"    "White"    "Black"    "Asian"    "Hispanic"
#[7] "White"    "White"    "White"    "Black"  

Here, I assumed that there is a single 1 per each row. If there are multiple 1s, we can use the option ties.method='first' or ties.method='last'.

---

Or another option is doing the %*% with the sequence of columns, subset the column names, and remove the prefix with sub.

 sub('[^.]+\\.', '', names(dat)[(as.matrix(dat) %*%seq_along(dat))[,1]])

---

Or we can use pmax

sub('[^.]+\\.', '', names(dat)[do.call(pmax,dat*seq_along(dat)[col(dat)])])