Generates a comprehensive data dictionary from DHS raw datasets, extracting variable names, labels, types, unique value counts, and missing data percentages. This function is particularly useful for exploring and documenting DHS survey datasets.
Value
A tibble with the following columns:
- var_name
Character. Variable name as it appears in the dataset
- var_label
Character. Human-readable label from the variable's label attribute, or empty string if no label exists
- var_type
Character. R data type(s) of the variable, comma-separated if multiple classes
- n_unique
Integer. Number of unique non-missing values
- pct_missing
Numeric. Percentage of missing values, rounded to 2 decimal places
Details
This function is designed to work with labeled data typically found in DHS datasets imported from SPSS or Stata files using the haven package. It safely handles variables without labels and provides a quick overview of data quality and structure.
The function extracts:
Variable labels from the "label" attribute
Data types using the class() function
Unique value counts excluding NA values
Missing data percentages as a quality metric
See also
dhs_read for loading DHS parquet datasets
Examples
if (FALSE) { # \dontrun{
# Load a DHS dataset
pr_data <- dhs_read(
path = dhs_data_path("01_data/parquet"),
file_type = "PR",
country_code = "KE",
survey_year = 2022
)
# Create data dictionary
dict <- make_dhs_raw_dictionary(pr_data)
# View first few entries
head(dict)
# Filter to see variables with high missing rates
dict |>
dplyr::filter(pct_missing > 50) |>
dplyr::arrange(desc(pct_missing))
# Find malaria-related variables
dict |>
dplyr::filter(grepl("malaria|fever|net", var_label, ignore.case = TRUE))
} # }