Convert Boolean indicator columns to a single factor column

Question

A similar questions was asked a few years ago here.

My setting is a little different. My indicator variables are not \"true\" dummy variables because they overlap. <

Answer 1

Here's one way using tidyverse functions

library(tibble)
library(dplyr)
library(tidyr)
dat %>% 
  rowid_to_column() %>% # keep data for each row together
  gather("col", "val", -rowid) %>% 
  mutate(rowid=factor(rowid)) %>% 
  filter(val==1) %>% 
  group_by(rowid) %>% 
  summarize(desired=paste(col, collapse=",")) %>%  #collapse values
  complete(rowid, fill = list(desired="none")) # add "none" for empty groups

#   rowid desired
#   <fct> <chr>  
# 1 1     none   
# 2 2     a      
# 3 3     a,b,c  

The basic idea involves reshaping the data so we can run functions of groups rather than running functions over rows of a data.frame which isn't as easy.