2

I need to change the reference level for a, b, d, and e variables. The reference level should be 3 for variable a and 2 for variable b, d and e. This can be done separately. But want to do at once or using for loop. Following is the code that I have written. But, showed an error message "'relevel' only for (unordered) factors".

tr <- data.frame(a = c(3, 2, 3, 3, 1), b = c(4, 3, 2, 2, 1), c = c(1, 2, 3, 2, 1),
                 d = c(1, 3, 1, 2, 2), e = c(1, 2, 0, 1, 2))

#convert to factor type
tr <- tr %>% mutate_if(is.numeric, as.factor)

# Relevel factors
col_set <- list(1, c(2, 4, 5))  # column index to relevel
r <- c("3", "2")      # referencing level

for (i in seq_along(col_set)){
  tr[col_set[[i]]] <- relevel(tr[col_set[[i]]], r[i])
}

**Error in relevel.default(tr[col_set[[i]]], r[i]) : 
  'relevel' only for (unordered) factors**
SAN
  • 25
  • 4

2 Answers2

1

You haven't quite got the indexing right. Remember tr[col_set[[i]]] may return a data frame (for example when i == 2), so you can't reorder it. You need to pick out the individual columns and reorder them. This requires an inner loop.

library(dplyr)

tr <- data.frame(a = c(3, 2, 3, 3, 1), b = c(4, 3, 2, 2, 1), c = c(1, 2, 3, 2, 1),
                 d = c(1, 3, 1, 2, 2), e = c(1, 2, 0, 1, 2))

tr <- tr %>% mutate_if(is.numeric, as.factor)

col_set <- list(1, c(2, 4, 5))  # column index to relevel
r <- c("3", "2")      # referencing level

for (i in seq_along(col_set))
  for(j in seq_along(tr[col_set[[i]]]))
    tr[col_set[[i]]][[j]] <-  relevel(tr[col_set[[i]]][[j]], r[i])

The result looks like this:

tr
#>   a b c d e
#> 1 3 4 1 1 1
#> 2 2 3 2 3 2
#> 3 3 2 3 1 0
#> 4 3 2 2 2 1
#> 5 1 1 1 2 2

and we can check the factor levels like this:

as.list(tr)
#> $a
#> [1] 3 2 3 3 1
#> Levels: 3 1 2
#> 
#> $b
#> [1] 4 3 2 2 1
#> Levels: 2 1 3 4
#> 
#> $c
#> [1] 1 2 3 2 1
#> Levels: 1 2 3
#> 
#> $d
#> [1] 1 3 1 2 2
#> Levels: 2 1 3
#> 
#> $e
#> [1] 1 2 0 1 2
#> Levels: 2 0 1

Created on 2020-06-15 by the reprex package (v0.3.0)

Allan Cameron
  • 147,086
  • 7
  • 49
  • 87
  • @ Allan Thanks for this. I'm not very familiar with indexing and I'm a new user for r. I think [[ can be used instead of $. Could you please explain the difference between tr[[i]] and tr[[i]][[j]]? Thanks once again – SAN Jun 16 '20 at 10:47
1

Instead of the for loops you can use Map to use relevel on specific columns.

tr[unlist(col_set)] <- Map(relevel, tr[unlist(col_set)], rep(r, lengths(col_set)))
str(tr)
#'data.frame':   5 obs. of  5 variables:
# $ a: Factor w/ 3 levels "3","1","2": 1 3 1 1 2
# $ b: Factor w/ 4 levels "2","1","3","4": 4 3 1 1 2
# $ c: Factor w/ 3 levels "1","2","3": 1 2 3 2 1
# $ d: Factor w/ 3 levels "2","1","3": 2 3 2 1 1
# $ e: Factor w/ 3 levels "2","0","1": 3 1 2 3 1
GKi
  • 37,245
  • 2
  • 26
  • 48
  • @ GKi Thanks for this one! It seems that the solution was simple. Can we use map2_df() instead of the map()? I tried it and unfortunately received an error msg. – SAN Jun 16 '20 at 10:42
  • I don't have the function `map2_df() `. – GKi Jun 16 '20 at 10:44