Derived Variables

Users can define variables that are derived from existing questionnaire responses in the dataset-specific config file, under a derived YAML block. See the following example, and the explanation below:

variables:
  HW001:
     name: "example variable"
     type: categorical
     levels:
       lvl1:
         name: "male"
         alternate_patterns:
           - "m"
       lvl2:
         name: "female"
       lvl3:
         name: "prefer not to answer"
derived:
  HW001_derived:
    name: "sex binary encoding"
    type: numeric
    code: |
      binary.sex <- rep(NA, length(HW001))
      is.male <- HW001 == "male" & !is.na(HW001)
      is.female <- HW001 == "female" & !is.na(HW001)
      binary.sex[is.male] <- 0
      binary.sex[is.female] <- 1
      binary.sex

Note that there are two main blocks in the config - variables, which contains the questionnaire response data, and derived, which contains new variables created from the questionnaire response data. The variables section is explained in more detail here. In the derived section, the one section that isn’t present in the variables block is the code section.

• Within the code block, you can write arbitrary R syntax. You have access to the variable names (e.g. HW001) from both the variables and the derived sections. The variable name gives you access to the entire vector of data, and you can use that and modify it, but this will not change the underlying original data

• The code block should return the value that should be stored as the derived variable (the vector binary.sex in the example above)

• The return value should be of length n.subjects or a single value, in which case it will be extended to be assigned to all subjects