Skip to contents

Creates a binary (1/0/NA) dummy variable indicating whether a household's first-ranked (main) food source is humanitarian assistance. The first-ranked source is taken from the ranked food-source question (fsl_source_food_ranked), whose value is an ordered, space-separated list of choice names.

Usage

add_food_source_assistance_d(
  df,
  source_food = "fsl_source_food_ranked",
  assistance = c("assistance_in_kind", "assistance_cva"),
  non_assistance = c("own_production", "purchase_cash", "purchase_credit", "hunting",
    "gathering", "exchange", "borrow", "gift", "begging"),
  undefined = c("other", "dnk", "pnta"),
  rank_sep = " "
)

Arguments

df

A data frame of household-level data.

source_food

Column name for the ranked food-source question. Its value is an ordered, space-separated list of choice names; the first entry is the household's main food source.

assistance

Character vector of choice names for humanitarian assistance food sources. The dummy is 1 when the first-ranked source is in this set. Must not overlap non_assistance or undefined.

non_assistance

Character vector of known, valid non-assistance food sources. The dummy is 0 when the first-ranked source is in this set. Must not overlap assistance or undefined.

undefined

Character vector of non-substantive responses (e.g. other, dnk, pnta). The dummy is NA when the first-ranked source is in this set, and also when it is missing.

rank_sep

Separator between ranked choices in source_food. Default " ".

Value

A data frame with additional columns:

  • fsl_food_source_main: The first-ranked (main) food source, or NA when missing.

  • fsl_food_source_assistance_d: 1 if the first-ranked food source is humanitarian assistance; 0 if it is a known non-assistance source; NA if it cannot be classified.

Details

ANA function 2025 INDICATOR ID: IND052 2026 METRIC ID: TBD

Examples

df <- data.frame(
  fsl_source_food_ranked = c(
    "assistance_in_kind purchase_cash",
    "purchase_cash own_production",
    "other purchase_cash",
    NA
  )
)
add_food_source_assistance_d(df)
#>             fsl_source_food_ranked fsl_food_source_main
#> 1 assistance_in_kind purchase_cash   assistance_in_kind
#> 2     purchase_cash own_production        purchase_cash
#> 3              other purchase_cash                other
#> 4                             <NA>                 <NA>
#>   fsl_food_source_assistance_d
#> 1                            1
#> 2                            0
#> 3                           NA
#> 4                           NA