Branching items

A branching item in an ESM survey is a type of question that presents respondents with specific propositions (case 1) or a specific follow-up set of items (case 2) based on their initial response. However, it is important to ensure that participants’ subsequent responses align with the branching structure of the survey.

Case 1: Follow-up answer choices

In our data example, in the function of the answer at the branching variable (“branching_var”) only specific choices of the following item (“branching_item”) are proposed to the participant, as follows:

Branch_var	Branch_item
repA	choiceA ; choiceB
repB	choiceC ; choiceD

Coherence (i.e., repA -> choiceC) must be checked. To this end, we can count the number of occurrences between “branching_var” and “branching_item” values:

data %>% 
    group_by(branch_var, branch_item) %>% 
    summarise(n())

# A tibble: 7 × 3
# Groups:   branch_var [2]
  branch_var branch_item `n()`
  <chr>      <chr>       <int>
1 repA       choiceA       600
2 repA       choiceB       584
3 repA       choiceD         2
4 repB       choiceA         3
5 repB       choiceB         6
6 repB       choiceC      1531
7 repB       choiceD      1474

The function highlights that occurrences of incoherent combinations of values exists (e.g., repA -> choiceC, repB -> choiceA).

Solving this issue: if there is no way to get the correct values, you might consider to handle this issue by setting the incoherent values of the incoherent conditions as missing.

# Define the incoherent combinations.
incoherent = list(c("repA", "choiceC"), c("repA", "choiceD"),
                  c("repB", "choiceA"), c("repB", "choiceB")) 

# Whenever incoherent, set as missing the values.
for (inc in incoherent){
    condition_ = data$branch_var==inc[1] & data$branch_item==inc[2] & !is.na(data$branch_var) & !is.na(data$branch_item)
    data[condition_, "branch_var"] = NA
    data[condition_, "branch_item"] = NA
}

Don’t forget to check that the issues have been solved using the previous R code.

Case 2: Follow-up items

In our data example, in the function of the answer at the “branching_var” only specific variables are presented to the participants, as follows:

Branch_var	Items
repA	‘PA1’
repB	‘PA2’

Again coherence (i.e., repA -> ‘PA1’, repB -> ‘PA2’) must be checked. To this end, we count the number of occurrences between non-missing values in ‘PA1’ and ‘PA2’ and the ‘branch_var’ values:

data %>% 
    mutate(PA1_NA= !is.na(PA1),
           PA2_NA= !is.na(PA2)) %>%
    group_by(branch_var, PA1_NA, PA2_NA) %>% 
    summarise(n())

# A tibble: 6 × 4
# Groups:   branch_var, PA1_NA [4]
  branch_var PA1_NA PA2_NA `n()`
  <chr>      <lgl>  <lgl>  <int>
1 repA       FALSE  FALSE    351
2 repA       TRUE   FALSE    881
3 repA       TRUE   TRUE       9
4 repB       FALSE  FALSE    903
5 repB       FALSE  TRUE    2049
6 repB       TRUE   TRUE       7

Solving this issue: one solution is to assign missing values to the follow-up items (‘PA1’, ‘PA2’) whenever an observation have non-coherent values.

data[data$branch_var=="repA" & !is.na(data$PA2), "PA2"] = NA
data[data$branch_var=="repB" & !is.na(data$PA1), "PA1"] = NA