Creating new column informing how many unique values there are in multiple columns

So, I have a dataframe that looks like this, but with 10k rows:

If my data had only one student, using the following command:

dep_edu2016_4x = data.frame(table(unlist(dep_edu2016_4x)))

Would tell me how many times this student changed schools in a year through the number of rows the table would return, minus the studentID row. I’ve already made a dummy to know if they’ve changed schools, now I need a solution to create a new column informing me how many times each student changed schools

>Solution :

dep_edu2016_4x$unique_schools <- apply(
  subset(dep_edu2016_4x, select = school_ID_1:school_ID_4), 1,
  function(z) length(unique(z)) - 1L)
dep_edu2016_4x
#   studentID school_ID_1 school_ID_2 school_ID_3 school_ID_4 unique_schools
# 1      5555        3321        3311        3311        3311              1
# 2      5544        3347        3326        3350        3347              2
# 3      5533        3364        3314        3328        3302              3
# 4      5566        3389        3351        3379        3384              3
# 5      5577        3312        3303        3320        3320              2

Walk-through:

• subset(...) is one way to extract just the columns we need, including a range of columns; we could easily have started with dep_edu2016_4x[,-1] to select all except the studentID, but (1) I don’t assume there are no other columns; and even if this is your whole dataset, then (2) once we add that column, we cannot recalculate it without biasing it with that additional column.

• apply(X, MARGIN, FUN) operates on each row (MARGIN=1) of the dataset (X=x), and applies the function (FUN) to that data. In this case, for each row it returns a single integer.

• length(unique(z)) - 1L calculates the number of distinct schools minus 1, a good proxy for the number of changes.

The count of school-changes is slightly incorrect, though: if a student starts at school 3321, changes to 3311, then changes back to 3321, then length(unique(.))-1L will not work. (This appears to be the case for student 5544.) We’ll need something a little more robust if you want to guard against that:

changes <- function(z) if (length(z)) sum(z[-1] != z[-length(z)]) else 0L
dep_edu2016_4x$schools_changes <- apply(subset(dep_edu2016_4x, select = school_ID_1:school_ID_4), 1, changes)
dep_edu2016_4x
#   studentID school_ID_1 school_ID_2 school_ID_3 school_ID_4 unique_schools schools_changes
# 1      5555        3321        3311        3311        3311              1               1
# 2      5544        3347        3326        3350        3347              2               3
# 3      5533        3364        3314        3328        3302              3               3
# 4      5566        3389        3351        3379        3384              3               3
# 5      5577        3312        3303        3320        3320              2               2