Issue with generated values

[reikookamoto]

Reposting this issue here as per @DougManuel's suggestion.

Issue pertains to code currently sitting in the package-repatriation branch.

When we use the prop_NA argument in create_con_var() (as shown below) and one of the recStart to recEnd relationship for the variable in question is else --> NA::b (also shown below), the new column in the output data frame contains "else".

Should NA::b appear in the output instead of "else"?

mock_drinks_norm <- create_con_var(
    var_raw = "alcdwky",
    cycle = "cycle1",   
    variable_details = variable_details,
    length = 1000,
    df_mock = data.frame(),
    prop_NA = 0.1,
    distribution = "normal"
  )

[DougManuel]

What should happen

The generated mock data should contain:

• Explicit missing codes from metadata (like 996, 997-999) where specified
• Actual R NA values for the else rule (not the literal string "else")

Why

We're generating raw/source data (before harmonization), not harmonized output.

The else → NA::b mapping is a rule, not a value that appears in raw data. In real surveys:

• Standard missing codes (996-999) appear as those actual numbers
• Unexpected/garbage values (blanks, typos, age 127) trigger the else rule during harmonization
• In our mock data, we should represent "unexpected values" as NA since we can't predict what garbage might appear

CCHS has high data quality, but there are always occasional issues - one respondent had age 127 (oldest person ever!). Eventually it would be fun to generate realistic garbage to test else rules thoroughly, but for now, generating NA is what we need.

Context for debugging

The issue is in how get_variable_categories() handles the else special case when collecting NA codes. The function returns recStart values, including the literal "else" string, which then gets sampled into the output.

Consider: should else be skipped when building the list of NA codes, or handled differently? The existing logic in create_con_var() already has a fallback to generate actual NA if the NA codes list is empty.

What we also need

We need more documentation and context about what data we are creating. That is, "raw" data before harmonization vs harmonized output. The recStart → recEnd mappings describe transformations applied during harmonization, not the structure of the mock data itself. This distinction should be clearer in our documentation so the intent of MockData is explicit.

[DougManuel]

Addressed in PR #9