Package {retroharmonize}

Type:	Package
Title:	Ex Post Survey Data Harmonization
Version:	0.2.8
Date:	2026-05-19
Maintainer:	Daniel Antal <daniel.antal@dataobservatory.eu>
Description:	Assist in reproducible retrospective (ex-post) harmonization of data, particularly individual level survey data, by providing tools for organizing metadata, standardizing the coding of variables, and variable names and value labels, including missing values, and documenting the data transformations, with the help of comprehensive s3 classes.
License:	GPL-3
URL:	https://retroharmonize.dataobservatory.eu/
BugReports:	https://github.com/dataobservatory-eu/retroharmonize/issues
Depends:	R (≥ 3.5.0)
Imports:	assertthat, cli, dataset, dplyr (≥ 1.0.0), fs, glue, haven, here, labelled, magrittr, purrr, rlang, snakecase, stats, stringr, tibble, tidyr, tidyselect, utils, vctrs
Suggests:	covr, ggplot2, knitr, markdown, png, rmarkdown, pillar, spelling, statcodelists, testthat (≥ 3.0.0)
VignetteBuilder:	knitr
Config/testthat/edition:	3
Encoding:	UTF-8
Language:	en-US
Config/roxygen2/version:	8.0.0
NeedsCompilation:	no
Packaged:	2026-05-20 19:04:59 UTC; DanielAntal
Author:	Daniel Antal [aut, cre], Marta Kolczynska [ctb]
Repository:	CRAN
Date/Publication:	2026-05-21 07:20:09 UTC

retroharmonize: Ex Post Survey Data Harmonization

Description

Assist in reproducible retrospective (ex-post) harmonization of data, particularly individual level survey data, by providing tools for organizing metadata, standardizing the coding of variables, and variable names and value labels, including missing values, and documenting the data transformations, with the help of comprehensive s3 classes.

Author(s)

Maintainer: Daniel Antal daniel.antal@dataobservatory.eu (ORCID)

Authors:

• Daniel Antal daniel.antal@dataobservatory.eu (ORCID)

Other contributors:

• Marta Kolczynska mkolczynska@gmail.com (ORCID) [contributor]

See Also

Useful links:

• https://retroharmonize.dataobservatory.eu/

• Report bugs at https://github.com/dataobservatory-eu/retroharmonize/issues

Pipe operator

Description

See magrittr::%>% for details.

Usage

lhs %>% rhs

Labelled to labelled_spss_survey

Description

Labelled to labelled_spss_survey

Usage

as_labelled_spss_survey(x, id)

Arguments

Value

A vector of labelled_spss_survey

See Also

Other type conversion functions: labelled_spss_survey_coercion

Collect labels from metadata file

Description

Collect labels from metadata file

Usage

collect_val_labels(metadata)

collect_na_labels(metadata)

Arguments

Value

The unique valid labels or the user-defined missing labels found in all the files analyzed in metadata.

See Also

Other harmonization functions: crosswalk_surveys(), harmonize_na_values(), harmonize_survey_values(), harmonize_values(), harmonize_var_names(), is.crosswalk_table(), label_normalize()

Examples

test_survey <- retroharmonize::read_rds(
  file = system.file("examples", "ZA7576.rds",
    package = "retroharmonize"
  ),
  id = "test"
)
example_metadata <- metadata_create(test_survey)

collect_val_labels(metadata = example_metadata)
collect_na_labels(metadata = example_metadata)

Concatenate haven_labelled_spss vectors

Description

Concatenate haven_labelled_spss vectors

Usage

concatenate(x, y)

Arguments

Value

A concatenated haven_labelled_spss vector. Returns an error if the attributes do not match. Gives a warning when only the variable label do not match.

Examples

v1 <- labelled::labelled(
  c(3, 4, 4, 3, 8, 9),
  c(YES = 3, NO = 4, `WRONG LABEL` = 8, REFUSED = 9)
)
v2 <- labelled::labelled(
  c(4, 3, 3, 9),
  c(YES = 3, NO = 4, `WRONG LABEL` = 8, REFUSED = 9)
)
s1 <- haven::labelled_spss(
  x = unclass(v1), # remove labels from earlier defined
  labels = labelled::val_labels(v1), # use the labels from earlier defined
  na_values = NULL,
  na_range = 8:9,
  label = "Variable Example"
)

s2 <- haven::labelled_spss(
  x = unclass(v2), # remove labels from earlier defined
  labels = labelled::val_labels(v2), # use the labels from earlier defined
  na_values = NULL,
  na_range = 8:9,
  label = "Variable Example"
)
concatenate(s1, s2)

Convert to haven_labelled_spss

Description

Convert to haven_labelled_spss

Usage

convert_to_labelled_spss(x, na_labels = NULL)

Arguments

Value

A haven_labelled_spss vector

Create a survey codebook

Description

Expand survey metadata into a long-format codebook of value labels.

Usage

create_codebook(metadata = NULL, survey = NULL)

codebook_waves_create(waves)

codebook_surveys_create(survey_list)

Arguments

Details

'create_codebook()' takes survey-level metadata and returns a tidy data frame describing all labelled variables and their associated value labels. Each row corresponds to a single value label, classified as either a valid value or a missing value.

Unlabelled numeric and character variables are excluded.

For multiple survey waves, use [codebook_surveys_create()].

If both 'metadata' and 'survey' are provided, 'survey' takes precedence.

Value

A data frame with one row per value label, including:

• survey identifiers ('id', 'filename')

• original variable names and labels

• value codes and value labels

• label type ('"valid"' or '"missing"')

• summary counts of labels

Additional user-defined metadata columns present in the input metadata are preserved.

See Also

[metadata_create()], [codebook_surveys_create()]

Other metadata functions: is.crosswalk_table(), metadata_create(), metadata_survey_create()

Examples

survey <- read_rds(
  system.file("examples", "ZA7576.rds", package = "retroharmonize")
)

cb <- create_codebook(survey = survey)
head(cb)


examples_dir <- system.file("examples", package = "retroharmonize")
survey_list <- dir(examples_dir)[grepl("\\.rds", dir(examples_dir))]

example_surveys <- read_surveys(
  file.path(examples_dir, survey_list),
  save_to_rds = FALSE
)

codebook_surveys_create(example_surveys)

Crosswalk and harmonize surveys

Description

Harmonize one or more surveys using a crosswalk table that defines how variable names, value labels, numeric codes, and variable classes should be aligned across surveys.

Usage

crosswalk_surveys(
  crosswalk_table,
  survey_list = NULL,
  survey_paths = NULL,
  import_path = NULL,
  na_values = NULL
)

crosswalk(survey_list, crosswalk_table, na_values = NULL)

Arguments

Details

A crosswalk table can be created with [crosswalk_table_create()] or supplied manually as a data frame. At a minimum, the table must contain columns 'id', 'var_name_orig', and 'var_name_target'. Additional columns enable harmonization of value labels, numeric codes, missing values, and variable classes.

Value

'crosswalk_surveys()' returns a list of harmonized survey data frames. 'crosswalk()' returns either a single data frame (if only one survey is harmonized) or a merged data frame combining all harmonized surveys.

See Also

[crosswalk_table_create()] to create a crosswalk table, [harmonize_survey_variables()] for lower-level variable harmonization.

Other harmonization functions: collect_val_labels(), harmonize_na_values(), harmonize_survey_values(), harmonize_values(), harmonize_var_names(), is.crosswalk_table(), label_normalize()

Examples

## Not run: 
examples_dir <- system.file("examples", package = "retroharmonize")
survey_files <- dir(examples_dir, pattern = "\\.rds$")

surveys <- read_surveys(
  file.path(examples_dir, survey_files),
  save_to_rds = FALSE
)

metadata <- metadata_create(survey_list = surveys)

crosswalk_table <- crosswalk_table_create(metadata)

harmonized <- crosswalk_surveys(
  crosswalk_table = crosswalk_table,
  survey_list = surveys
)

## End(Not run)

Document survey item provenance

Description

Document the current and historical coding, labels, missing values, and survey provenance of a harmonized survey variable.

Usage

document_survey_item(x)

Arguments

Value

A named list containing:

• current and historical value coding,

• variable labels,

• valid and missing value definitions,

• original variable names,

• survey identifiers.

See Also

Other documentation functions: document_surveys()

Examples

var1 <- labelled::labelled_spss(
  x = c(1, 0, 1, 1, 0, 8, 9),
  labels = c(
    "TRUST" = 1,
    "NOT TRUST" = 0,
    "DON'T KNOW" = 8,
    "INAP. HERE" = 9
  ),
  na_values = c(8, 9)
)

var2 <- labelled::labelled_spss(
  x = c(2, 2, 8, 9, 1, 1),
  labels = c(
    "Tend to trust" = 1,
    "Tend not to trust" = 2,
    "DK" = 8,
    "Inap" = 9
  ),
  na_values = c(8, 9)
)

harmonization <- list(
  from = c(
    "^tend\\sto|^trust",
    "^tend\\snot|not\\strust",
    "^dk|^don",
    "^inap"
  ),
  to = c(
    "trust",
    "not_trust",
    "do_not_know",
    "inap"
  ),
  numeric_values = c(1, 0, 99997, 99999)
)

missing_values <- c(
  "do_not_know" = 99997,
  "inap" = 99999
)

h1 <- harmonize_values(
  x = var1,
  harmonize_label = "Do you trust the European Union?",
  harmonize_labels = harmonization,
  na_values = missing_values,
  id = "survey1"
)

h2 <- harmonize_values(
  x = var2,
  harmonize_label = "Do you trust the European Union?",
  harmonize_labels = harmonization,
  na_values = missing_values,
  id = "survey2"
)

h3 <- concatenate(h1, h2)

document_survey_item(h3)

Document survey lists

Description

Document the key attributes surveys in a survey list.

Usage

document_surveys(survey_list = NULL, survey_paths = NULL, .f = NULL)

document_waves(waves)

Arguments

Details

The function has two alternative input parameters. If survey_list is the input, it returns the name of the original source data file, the number of rows and columns, and the size of the object as stored in memory. In case survey_paths contains the source data files, it will sequentially read those files, and add the file size, the last access and the last modified time attributes.

The earlier form document_waves is deprecated. Currently called document_surveys.

Value

Returns a data frame with the key attributes of the surveys in a survey list: the name of the data file, the number of rows and columns, and the size of the object as stored in memory.

See Also

Other documentation functions: document_survey_item()

Examples

examples_dir <- system.file("examples", package = "retroharmonize")

my_rds_files <- dir(examples_dir)[grepl(
  ".rds",
  dir(examples_dir)
)]

example_surveys <- read_surveys(file.path(examples_dir, my_rds_files))

documented <- document_surveys(example_surveys)

attr(documented, "original_list")
documented

document_surveys(survey_paths = file.path(examples_dir, my_rds_files))

Find import function by file extension

Description

This is an internal utility to select the appropriate importing function.

Usage

find_import_function(file_path)

Value

The name of the function that should read file_path based on the file extension.

Harmonize na_values in haven_labelled_spss

Description

Harmonize na_values in haven_labelled_spss

Usage

harmonize_na_values(df)

Arguments

Value

A tibble where the na_values are consistent

See Also

Other harmonization functions: collect_val_labels(), crosswalk_surveys(), harmonize_survey_values(), harmonize_values(), harmonize_var_names(), is.crosswalk_table(), label_normalize()

Examples

examples_dir <- system.file(
  "examples",
  package = "retroharmonize"
)

test_read <- read_rds(
  file.path(examples_dir, "ZA7576.rds"),
  id = "ZA7576",
  doi = "test_doi"
)

harmonize_na_values(test_read)

Harmonize values in surveys

Description

Harmonize value codes and value labels across multiple surveys and combine them into a single data frame.

Usage

harmonize_survey_values(survey_list, .f, status_message = FALSE)

harmonize_waves(waves, .f, status_message = FALSE)

Arguments

Details

The function first aligns the structure of all surveys by ensuring that they contain the same set of variables. Missing variables are added and filled with appropriate missing values depending on their type.

Variables of class "retroharmonize_labelled_spss_survey" are then harmonized by applying a user-supplied function .f to each variable separately within each survey.

The harmonization function .f must return a vector of the same length as its input. If .f returns NULL, the original variable is kept unchanged.

Prior to version 0.2.0 this function was called harmonize_waves.

The earlier form harmonize_waves is deprecated. The function is currently called harmonize_waves.

Value

A data frame containing the row-wise combination of all surveys, with harmonized labelled variables and preserved attributes describing the original surveys.

See Also

Other harmonization functions: collect_val_labels(), crosswalk_surveys(), harmonize_na_values(), harmonize_values(), harmonize_var_names(), is.crosswalk_table(), label_normalize()

Examples

examples_dir <- system.file("examples", package = "retroharmonize")
survey_files <- dir(examples_dir, pattern = "\\.rds$", full.names = TRUE)

surveys <- read_surveys(
  survey_files,
  export_path = NULL
)

# Keep only supported variable types
surveys <- lapply(
  surveys,
  function(s) {
    s[, vapply(
      s,
      function(x) {
        inherits(x, c(
          "retroharmonize_labelled_spss_survey",
          "numeric",
          "character",
          "Date"
        ))
      },
      logical(1)
    )]
  }
)

# Identity harmonization (no-op)
harmonized <- harmonize_survey_values(
  survey_list = surveys,
  .f = function(x) x,
  status_message = FALSE
)

head(harmonized)

Read a survey from a CSV file

Description

Import a survey stored in a CSV file and return it as a survey object with attached dataset- and survey-level metadata.

Usage

harmonize_survey_variables(
  crosswalk_table,
  subset_name = "subset",
  survey_list = NULL,
  survey_paths = NULL,
  import_path = NULL,
  export_path = NULL
)

Arguments

Details

The CSV file is read using [utils::read.csv()]. Character variables with more than one unique value are automatically converted to labelled factors. A unique row identifier is added and labelled.

If the file cannot be read, an empty survey object is returned with a warning.

If a column named '"X"' is present (commonly created by 'write.csv()'), it is removed automatically.

Value

An object of class '"survey"', which is a data frame with attached survey- and dataset-level metadata.

See Also

[read_rds()] for importing surveys from RDS files, [survey_df()] for constructing survey objects manually.

Other import functions: pull_survey(), read_csv(), read_dta(), read_rds(), read_spss(), read_surveys()

Examples

# Create a temporary CSV file from an example survey
path <- system.file("examples", "ZA7576.rds",
  package = "retroharmonize"
)
survey <- read_rds(path)

tmp <- tempfile(fileext = ".csv")
write.csv(survey, tmp, row.names = FALSE)

# Read the CSV file back as a survey
re_read <- read_csv(
  file = tmp,
  id = "ZA7576",
  doi = "10.0000/example"
)

Harmonize values and labels of labelled vectors

Description

'harmonize_values()' converts heterogeneous labelled survey vectors into a harmonized representation suitable for cross-survey integration.

The function:

- harmonizes value labels using regex-based matching; - assigns harmonized numeric codes; - preserves original coding metadata; - standardizes user-defined missing values; - preserves SPSS-style labelled metadata; - and records provenance attributes.

Usage

harmonize_values(
  x,
  harmonize_label = NULL,
  harmonize_labels = NULL,
  na_values = c(do_not_know = 99997, declined = 99998, inap = 99999),
  na_range = NULL,
  id = "survey_id",
  name_orig = NULL,
  remove = NULL,
  perl = FALSE
)

Arguments

Details

Create a harmonized labelled vector with standardized value labels, numeric coding, and missing value definitions.

Harmonization is performed using a harmonization table supplied via 'harmonize_labels'.

The harmonization table must contain:

- 'from': regex patterns matching original labels; - 'to': harmonized labels; - 'numeric_values': harmonized numeric codes.

Original labels and numeric codes are preserved in attributes attached to the returned vector.

If no harmonization table is supplied, the function still attempts to normalize common missing value labels such as:

- '"inap"' - '"declined"' - '"do_not_know"'

Value

A harmonized 'haven_labelled_spss' vector.

The returned vector preserves:

- harmonized value labels; - harmonized numeric coding; - SPSS missing value metadata; - original coding metadata; - survey provenance metadata.

See Also

[harmonize_var_names()]

Other harmonization functions: collect_val_labels(), crosswalk_surveys(), harmonize_na_values(), harmonize_survey_values(), harmonize_var_names(), is.crosswalk_table(), label_normalize()

Examples

var1 <- labelled::labelled_spss(
  x = c(1, 0, 1, 1, 0, 8, 9),
  labels = c(
    "TRUST" = 1,
    "NOT TRUST" = 0,
    "DON'T KNOW" = 8,
    "INAP. HERE" = 9
  ),
  na_values = c(8, 9)
)

harmonize_values(
  var1,
  harmonize_labels = list(
    from = c(
      "^tend\\sto|^trust",
      "^tend\\snot|not\\strust",
      "^dk|^don",
      "^inap"
    ),
    to = c(
      "trust",
      "not_trust",
      "do_not_know",
      "inap"
    ),
    numeric_values = c(
      1,
      0,
      99997,
      99999
    )
  ),
  na_values = c(
    "do_not_know" = 99997,
    "inap" = 99999
  ),
  id = "survey_id"
)

Harmonize variable names across surveys

Description

'harmonize_var_names()' renames variables across multiple surveys to a shared harmonized naming scheme.

The harmonization rules are defined in a metadata table, typically created with [metadata_create()].

Usage

harmonize_var_names(
  survey_list,
  metadata,
  old = "var_name_orig",
  new = "var_name_suggested",
  rowids = TRUE
)

Arguments

Details

Harmonize variable names in a list of survey objects using a metadata crosswalk table.

The function can also be used for survey subsetting workflows. If 'metadata' contains only a subset of variables for a survey, only those variables are retained in the harmonized output.

Value

A list of surveys with harmonized variable names.

See Also

[metadata_create()], [crosswalk()]

Other harmonization functions: collect_val_labels(), crosswalk_surveys(), harmonize_na_values(), harmonize_survey_values(), harmonize_values(), is.crosswalk_table(), label_normalize()

Examples

examples_dir <- system.file(
  "examples",
  package = "retroharmonize"
)

survey_files <- dir(
  examples_dir,
  pattern = "\\.rds$"
)

example_surveys <- read_surveys(
  file.path(examples_dir, survey_files)
)

metadata <- metadata_create(
  example_surveys
)

metadata$var_name_suggested <-
  label_normalize(metadata$var_name)

metadata$var_name_suggested[
  metadata$label_orig == "age_education"
] <- "age_education"

harmonized_surveys <- harmonize_var_names(
  survey_list = example_surveys,
  metadata = metadata
)

harmonized_surveys[[1]]

Here

Description

A utility to make sure the system files of the package and other files are always found, regardless if they are in an example or vignette context.

Details

See here::here for details.

Examples

dir(here("inst", "examples"))

Validate a crosswalk table

Description

Create a crosswalk table with the source variable names and variable labels.

Usage

is.crosswalk_table(ctable)

crosswalk_table_create(metadata)

Arguments

Details

The table contains a var_name_target and val_label_target column, but these values need to be set by further manual or reproducible harmonization steps.

Value

A tibble with raw crosswalk table. It contains all harmonization tasks, but the target values need to be set by further manipulations.

See Also

Other metadata functions: create_codebook(), metadata_create(), metadata_survey_create()

Other harmonization functions: collect_val_labels(), crosswalk_surveys(), harmonize_na_values(), harmonize_survey_values(), harmonize_values(), harmonize_var_names(), label_normalize()

Test whether missing values need harmonization

Description

Checks whether both 'na_values' and 'na_range' attributes are present on a labelled vector.

Usage

is.na_range_to_values(x)

Arguments

Value

Logical scalar.

Create a survey object

Description

Construct a survey object from a data frame or tibble by attaching survey-level metadata such as an identifier, source filename, and basic dataset-level descriptive metadata.

Usage

is.survey_df(x)

survey_df(
  x,
  title = NULL,
  creator = person("Unknown", "Creator"),
  dataset_bibentry = NULL,
  dataset_subject = NULL,
  identifier,
  filename
)

is.survey_df(x)

## S3 method for class 'survey_df'
print(x, ...)

Arguments

Details

This function is primarily intended for use by import helpers such as [read_rds()], [read_spss()], [read_dta()], and [read_csv()]. Most users will not need to call it directly.

Value

An object of class '"survey_df"', which is a data frame with additional survey-level metadata stored as attributes and dataset-level metadata stored using the 'dataset' package.

See Also

[read_survey()] for importing survey data from external files.

Other importing functions: survey()

Examples

survey_df(
  x = data.frame(
    rowid = 1:6,
    observations = runif(6)
  ),
  identifier = "example",
  filename = "no_file"
)

Normalize value and variable labels

Description

label_normalize removes special characters, whitespace, and other typical typing errors.

Usage

label_normalize(x)

var_label_normalize(x)

val_label_normalize(x)

Arguments

Details

var_label_normalize and val_label_normalize removes possible chunks from question identifiers.

The functions var_label_normalize and val_label_normalize may be differently implemented for various survey series.

Value

Returns a suggested, normalized label without special characters. The var_label_normalize and val_label_normalize returns them in snake_case for programmatic use.

See Also

Other variable label harmonization functions: na_range_to_values()

Other harmonization functions: collect_val_labels(), crosswalk_surveys(), harmonize_na_values(), harmonize_survey_values(), harmonize_values(), harmonize_var_names(), is.crosswalk_table()

Examples

label_normalize(
  c(
    "Don't know", " TRUST", "DO NOT  TRUST",
    "inap in Q.3", "Not 100%", "TRUST < 50%",
    "TRUST >=90%", "Verify & Check", "TRUST 99%+"
  )
)

var_label_normalize(
  c(
    "Q1_Do you trust the national government?",
    " Do you trust the European Commission"
  )
)

val_label_normalize(
  c(
    "Q1_Do you trust the national government?",
    " Do you trust the European Commission"
  )
)

Coercion methods for labelled survey vectors

Description

Convert labelled SPSS-style survey vectors to common R data types. These helpers provide consistent coercion behavior for '"retroharmonize_labelled_spss_survey"' objects while respecting labelled missing values.

Usage

as_numeric(x)

as_character(x)

as_factor(x, levels = "default", ordered = FALSE)

Arguments

Value

* 'as_numeric()' returns a numeric vector with labelled missing values converted to 'NA'. * 'as_character()' returns a character vector based on the factor representation of 'x'. * 'as_factor()' returns a factor with levels derived from value labels.

See Also

[labelled_spss_survey()], [haven::as_factor()]

Other type conversion functions: as_labelled_spss_survey()

vctrs type and casting methods for labelled survey vectors

Description

These methods define how retroharmonize_labelled_spss_survey objects interact with base vectors and with each other in vctrs-based operations such as concatenation, binding, and coercion.

Usage

## S3 method for class 'retroharmonize_labelled_spss_survey.double'
vec_ptype2(x, y, ...)

## S3 method for class 'double.retroharmonize_labelled_spss_survey'
vec_ptype2(x, y, ...)

## S3 method for class 'integer.retroharmonize_labelled_spss_survey'
vec_ptype2(x, y, ...)

## S3 method for class 'retroharmonize_labelled_spss_survey.integer'
vec_ptype2(x, y, ...)

## S3 method for class 'double.retroharmonize_labelled_spss_survey'
vec_cast(x, to, ...)

## S3 method for class 'integer.retroharmonize_labelled_spss_survey'
vec_cast(x, to, ...)

## S3 method for class 'character.retroharmonize_labelled_spss_survey'
vec_cast(x, to, ...)

Details

They ensure that labelled survey vectors:

• combine safely with numeric vectors,

• cast consistently to base types,

• error on incompatible coercions.

These functions are part of the internal type system and are not intended to be called directly by users.

Merge and harmonize surveys

Description

'merge_surveys()' applies a harmonization specification to a list of survey objects and returns harmonized survey datasets with aligned variable names and metadata.

Usage

merge_surveys(survey_list, var_harmonization)

Arguments

Details

Harmonize variable names, labels, and identifiers across multiple surveys using a metadata crosswalk table.

Prior to version 0.2.0 this function was called 'merge_waves()', reflecting terminology commonly used in Eurobarometer surveys.

The harmonization table supplied in 'var_harmonization' typically originates from [metadata_create()] and contains mappings between original and harmonized variable names.

Value

A list of harmonized survey objects with standardized variable names and variable labels.

See Also

[metadata_create()]

Other survey harmonization functions: merge_waves()

Examples

examples_dir <- system.file(
  "examples",
  package = "retroharmonize"
)

survey_files <- dir(
  examples_dir,
  pattern = "\\.rds$",
  full.names = TRUE
)

example_surveys <- read_surveys(
  survey_files
)

metadata <- metadata_create(
  survey_list = example_surveys
)

to_harmonize <- metadata %>%
  dplyr::filter(
    var_name_orig %in% c("rowid", "w1") |
      grepl("^trust", var_label_orig)
  ) %>%
  dplyr::mutate(
    var_label = var_label_normalize(var_label_orig),
    var_name_target = val_label_normalize(var_label),
    var_name_target = ifelse(
      .data$var_name_orig %in%
        c("rowid", "w1", "wex"),
      .data$var_name_orig,
      .data$var_name_target
    )
  )

merged_surveys <- merge_surveys(
  survey_list = example_surveys,
  var_harmonization = to_harmonize
)

merged_surveys[[1]]

Deprecated wrapper for 'merge_surveys()'

Description

'merge_waves()' has been renamed to [merge_surveys()] for more general survey harmonization workflows.

Usage

merge_waves(waves, var_harmonization)

Arguments

Value

A list of harmonized survey objects.

See Also

[merge_surveys()]

Other survey harmonization functions: merge_surveys()

Create metadata tables from survey datasets

Description

Create a variable-level metadata table from one or more survey datasets. Metadata are extracted either from survey objects already loaded into memory or directly from survey files.

Usage

metadata_create(survey_list = NULL, survey_paths = NULL, .f = NULL)

metadata_waves_create(survey_list)

Arguments

Details

The resulting metadata table contains information about:

• variable names and labels,

• storage classes,

• value labels,

• user-defined missing values,

• and missing value ranges.

'metadata_create()' is a convenience wrapper around repeated [metadata_survey_create()] calls.

The form metadata_waves_create is deprecated.

Value

A data frame containing variable-level survey metadata.

See Also

[metadata_survey_create()], [create_variable_catalog()]

Other metadata functions: create_codebook(), is.crosswalk_table(), metadata_survey_create()

Examples

examples_dir <- system.file(
  "examples",
  package = "retroharmonize"
)

my_rds_files <- dir(examples_dir)[grepl(
  "\\.rds$",
  dir(examples_dir)
)]

example_surveys <- read_surveys(
  file.path(examples_dir, my_rds_files)
)

metadata_create(example_surveys)

Initialize a metadata data frame

Description

Initialize a metadata data frame

Usage

metadata_initialize(filename, id)

Arguments

Value

A nested data frame with metadata and the range of labels, na_values and the na_range itself.

Create variable-level metadata from a survey dataset

Description

Extract variable-level metadata from a survey dataset and return the result as a nested data frame.

Usage

metadata_survey_create(survey)

Arguments

Details

The metadata table contains:

• variable names and labels,

• imported storage classes,

• value labels,

• user-defined missing values,

• missing value ranges,

• and summary counts of labelled categories.

For multiple surveys, use [metadata_create()], which applies 'metadata_survey_create()' across a list of surveys or survey files.

Value

A nested data frame containing:

filename

Original survey file name.

id

Survey identifier.

var_name_orig

Original variable name.

class_orig

Imported storage class.

var_label_orig

Original variable label.

labels

List column of value labels.

valid_labels

List column of non-missing value labels.

na_labels

List column of user-defined missing labels.

na_range

List column containing user-defined missing ranges.

n_labels

Number of labelled categories.

n_valid_labels

Number of non-missing categories.

n_na_labels

Number of missing categories.

See Also

[metadata_create()], [create_variable_catalog()]

Other metadata functions: create_codebook(), is.crosswalk_table(), metadata_create()

Examples

metadata_survey_create(
  survey = read_rds(
    system.file(
      "examples",
      "ZA7576.rds",
      package = "retroharmonize"
    )
  )
)

Harmonize SPSS-style missing value ranges

Description

Ensure consistency between SPSS-style missing value ranges ('na_range') and explicit missing values ('na_values') for labelled survey vectors.

Usage

na_range_to_values(x)

Arguments

Details

When both attributes are present, this function:

• adjusts the missing range if it conflicts with existing missing values,

• derives missing values from the range when necessary,

• leaves non-SPSS-labelled vectors unchanged.

This harmonization is important before joining, binding, or summarizing survey data.

Value

The input vector with harmonized 'na_values' and 'na_range' attributes. If no harmonization is needed, 'x' is returned unchanged.

See Also

[labelled::na_range()], [labelled::na_values()], [as_numeric()]

Other variable label harmonization functions: label_normalize()

Labelled SPSS-style vectors with survey provenance

Description

Create a labelled vector compatible with [haven::labelled_spss()] that carries additional survey-level provenance metadata.

Usage

## S3 method for class 'retroharmonize_labelled_spss_survey'
print(x, ...)

labelled_spss_survey(
  x = double(),
  labels = NULL,
  na_values = NULL,
  na_range = NULL,
  label = NULL,
  id = NULL,
  name_orig = NULL
)

## S3 method for class 'retroharmonize_labelled_spss_survey'
x[i, ...]

## S3 method for class 'retroharmonize_labelled_spss_survey'
summary(object, ...)

## S3 replacement method for class 'retroharmonize_labelled_spss_survey'
names(x) <- value

## S3 method for class 'retroharmonize_labelled_spss_survey'
is.na(x)

## S3 method for class 'retroharmonize_labelled_spss_survey'
levels(x)

## S3 method for class 'retroharmonize_labelled_spss_survey'
format(x, ..., digits = getOption("digits"))

is.labelled_spss_survey(x)

## S3 method for class 'retroharmonize_labelled_spss_survey'
median(x, na.rm = TRUE, ...)

## S3 method for class 'retroharmonize_labelled_spss_survey'
quantile(x, probs, ...)

## S3 method for class 'retroharmonize_labelled_spss_survey'
weighted.mean(x, w, ...)

## S3 method for class 'retroharmonize_labelled_spss_survey'
mean(x, ...)

## S3 method for class 'retroharmonize_labelled_spss_survey'
sum(x, ...)

Arguments

Details

The resulting object behaves like a 'haven_labelled_spss' vector, but stores:

• a survey identifier;

• the original variable name;

• the original value coding.

Several arithmetic and statistical summary methods operate on the numeric representation of labelled survey vectors, converting SPSS-style missing values to 'NA' before computation.

You can coerce 'labelled_spss_survey' vectors to numeric, character or factor representation.

Value

An object of class '"retroharmonize_labelled_spss_survey"', extending [haven::labelled_spss()].

See Also

[haven::labelled_spss()], [as_factor()], [as_numeric()], [as_character()]

Examples

x <- labelled_spss_survey(
  x = c(1, 2, 9),
  labels = c(Yes = 1, No = 2),
  na_values = 9,
  id = "survey_1"
)

is.na(x)
as_factor(x)

Retrieve a survey from a survey list

Description

'pull_survey()' retrieves a survey object from a list created with [read_surveys()].

Surveys can be selected using:

- the survey identifier stored in the '"id"' attribute, or - the original source file name stored in the '"filename"' attribute.

Usage

pull_survey(survey_list, id = NULL, filename = NULL)

Arguments

Details

Extract a single 'survey' object from a list of surveys using either its survey identifier or source file name.

Either 'id' or 'filename' must be supplied.

The function throws an error if:

- neither argument is provided; - the requested survey cannot be found; - or multiple surveys match the query.

Value

A single 'survey' object.

See Also

[read_surveys()]

Other import functions: harmonize_survey_variables(), read_csv(), read_dta(), read_rds(), read_spss(), read_surveys()

Examples

examples_dir <- system.file(
  "examples",
  package = "retroharmonize"
)

survey_files <- dir(
  examples_dir,
  pattern = "\\.rds$"
)

example_surveys <- read_surveys(
  file.path(examples_dir, survey_files)
)

pull_survey(
  example_surveys,
  id = "ZA5913"
)

Data Input

Description

See utils::read.csv for details.

Read a survey dataset from a CSV file

Description

Import a survey dataset stored in comma-separated value ('.csv') format and convert it into a survey-compatible tibble with reproducibility metadata retained as attributes.

Usage

read_csv(file, id = NULL, doi = NULL, dataset_bibentry = NULL, ...)

Arguments

Details

The imported object is returned as a tibble with additional survey metadata such as identifiers, DOI references, and optional dataset bibliographic metadata.

Value

A tibble-like survey object with metadata attributes retained for reproducible workflows.

See Also

Other import functions: harmonize_survey_variables(), pull_survey(), read_dta(), read_rds(), read_spss(), read_surveys()

Examples

# Create a temporary CSV file:
path <- system.file(
  "examples",
  "ZA7576.rds",
  package = "retroharmonize"
)

read_survey <- read_rds(path)

test_csv_file <- tempfile(fileext = ".csv")

write.csv(
  x = read_survey,
  file = test_csv_file,
  row.names = FALSE
)

# Read the CSV file:
re_read <- read_csv(
  file = test_csv_file,
  id = "ZA7576",
  doi = "test_doi"
)

Read a Stata '.dta' survey file

Description

Import a survey dataset stored in Stata '.dta' format and convert it into a 'survey' object with harmonized metadata and labelled variables.

Usage

read_dta(file, id = NULL, doi = NULL, .name_repair = "unique")

Arguments

Details

This function wraps [haven::read_dta()] and adds:

- error handling, - survey metadata creation, - 'rowid' normalization, - preservation of variable labels, - conversion of labelled variables, - and provenance metadata.

Variable labels are preserved using the '"label"' attribute.

Labelled variables are converted to harmonized labelled survey vectors where possible. Variables that inherit from 'haven_labelled' but do not contain valid label definitions are converted back to standard vectors.

If the file cannot be read, the function returns an empty 'survey' object and emits a warning.

Value

A 'survey' object inheriting from 'data.frame' and 'tbl_df'.

See Also

Other import functions: harmonize_survey_variables(), pull_survey(), read_csv(), read_rds(), read_spss(), read_surveys()

Examples

path <- system.file(
  "examples",
  "iris.dta",
  package = "haven"
)

survey_object <- read_dta(path)

attr(survey_object, "id")
attr(survey_object, "filename")

Read a survey from an '.rds' file

Description

Import a serialized survey object stored in '.rds' format and return it as a 'survey' object with harmonized metadata attributes.

Usage

read_rds(file, dataset_bibentry = NULL, id = NULL, doi = NULL)

Arguments

Details

This function restores survey objects previously saved with [base::saveRDS()] or exported from the 'retroharmonize' workflow. The returned object retains survey metadata and gains additional provenance attributes such as source file name and file size.

If the file cannot be read, an empty 'survey' object is returned and a warning is emitted.

The function:

- restores the serialized object, - validates source file information, - normalizes 'rowid', - records provenance metadata, - and stores object and source file sizes as attributes.

Value

A 'survey' object inheriting from 'data.frame' and 'tbl_df' with survey metadata attributes.

See Also

Other import functions: harmonize_survey_variables(), pull_survey(), read_csv(), read_dta(), read_spss(), read_surveys()

Examples

path <- system.file(
  "examples",
  "ZA7576.rds",
  package = "retroharmonize"
)

survey_object <- read_rds(path)

attr(survey_object, "id")
attr(survey_object, "filename")
attr(survey_object, "doi")

Read SPSS survey files

Description

Import SPSS survey files in '.sav', '.zsav', or '.por' format and convert them into harmonized 'survey' objects with preserved metadata, labelled variables, and provenance information.

Usage

read_spss(
  file,
  user_na = TRUE,
  dataset_bibentry = NULL,
  id = NULL,
  doi = NULL,
  .name_repair = "unique"
)

Arguments

Details

This function wraps [haven::read_spss()] and adds:

- error handling, - harmonized survey metadata, - 'rowid' creation and normalization, - preservation of variable labels, - conversion of labelled SPSS vectors, - handling of malformed labelled variables, - and provenance metadata.

'read_sav()' reads both '.sav' and '.zsav' files. 'read_por()' reads portable SPSS '.por' files. 'read_spss()' automatically dispatches to the appropriate importer based on file extension.

Variables that inherit from 'haven_labelled' but do not contain valid label definitions are converted to standard numeric or character vectors.

If a file cannot be imported, the function returns an empty 'survey' object and emits a warning.

Value

A 'survey' object inheriting from 'data.frame' and 'tbl_df'.

Variable labels are stored in the '"label"' attribute of each variable.

Additional provenance metadata are stored as attributes, including:

- '"id"' - '"doi"' - '"object_size"' - '"source_file_size"'

See Also

Other import functions: harmonize_survey_variables(), pull_survey(), read_csv(), read_dta(), read_rds(), read_surveys()

Examples

path <- system.file(
  "examples",
  "iris.sav",
  package = "haven"
)

survey_object <- read_spss(path)

attr(survey_object, "id")
attr(survey_object, "filename")

Read survey files into memory or save as '.rds'

Description

Import one or more survey files into R using a consistent survey import workflow. The function supports SPSS ('.sav', '.por'), Stata ('.dta'), R ('.rds'), and CSV files.

Usage

read_surveys(
  survey_paths,
  .f = NULL,
  export_path = NULL,
  ids = NULL,
  dois = NULL,
  ...
)

read_survey(
  file_path,
  .f = NULL,
  export_path = NULL,
  doi = NULL,
  id = NULL,
  ...
)

Arguments

Details

Use [read_survey()] to import a single survey file and 'read_surveys()' to import multiple files in a loop.

When 'export_path' is 'NULL', imported surveys are returned as a list in memory. When 'export_path' is a valid directory, imported surveys are saved as '.rds' files with [base::saveRDS()].

Files that cannot be imported are skipped gracefully. A message is printed and 'NULL' is returned for the affected file.

Value

If 'export_path = NULL', a list of imported survey objects.

If 'export_path' is provided, a character vector containing exported '.rds' file names.

Imported surveys are returned as data frame-like [survey()] objects with metadata attributes retained for reproducible workflows.

See Also

[read_survey()], [survey()]

Other import functions: harmonize_survey_variables(), pull_survey(), read_csv(), read_dta(), read_rds(), read_spss()

Examples

file1 <- system.file(
  "examples",
  "ZA7576.rds",
  package = "retroharmonize"
)

file2 <- system.file(
  "examples",
  "ZA5913.rds",
  package = "retroharmonize"
)

surveys <- read_surveys(
  c(file1, file2),
  .f = "read_rds"
)

retroharmonize: Retrospective harmonization of survey data files

Description

The goal of retroharmonize is to facilitate retrospective (ex-post) harmonization of data, particularly survey data, in a reproducible manner. The package provides tools for organizing the metadata, standardizing the coding of variables, variable names and value labels, including missing values, and for documenting all transformations, with the help of comprehensive S3 classes.

import functions

Read data stored in formats with rich metadata, such as SPSS (.sav) files, and make them usable in a programmatic context.
read_spss: read an SPSS file and record metadata for reproducibility
read_rds: read an rds file and record metadata for reproducibility
read_surveys: programmatically read a list of surveys
pull_survey: pull a single survey from a survey list.

subsetting functions

subset_surveys: remove variables from surveys that cannot be harmonized.

variable name harmonization functions

harmonize_survey_variables: Create a list of surveys with harmonized variable names.

variable label harmonization functions

Create consistent coding and labelling.
harmonize_values: Harmonize the label list across surveys.
harmonize_survey_values: Create a list of surveys with harmonized value labels.
na_range_to_values: Make the na_range attributes, as imported from SPSS, consistent with the na_values attributes.
label_normalize removes special characters, whitespace, and other typical typing errors and helps the uniformization of labels and variable names.

survey harmonization functions

merge_surveys: Create a list of surveys with harmonized names and variable labels.
crosswalk_surveys: Create a list of surveys with harmonized variable names, harmonized value labels and harmonize R classes.
crosswalk: Create a joined data frame of surveys with harmonized variable names, harmonized value labels and harmonize R classes.

metadata functions

metadata_create: Createa metadata dataa from one or more survey.
metadata_survey_create: Create a joined metadata data frame from one survey.
create_codebook and codebook_waves_create crosswalk_table_create: Create an initial crosswalk table from a metadata data frame.

documentation functions

Make the workflow reproducible by recording the harmonization process. document_survey_item: Returns a list of the current and historic coding, labelling of the valid range and missing values or range, the history of the variable names and the history of the survey IDs. document_surveys: Document the key attributes surveys in a survey list.

type conversion functions

Consistently treat labels and SPSS-style user-defined missing values in the R language. survey helps constructing a valid survey data frame, and labelled_spss_survey helps creating a vector for a questionnaire item. as_numeric: convert to numeric values.
as_factor: convert to labels to factor levels.
as_character: convert to labels to characters.
as_labelled_spss_survey: convert labelled and labelled_spss vectors to labelled_spss_survey vectors.

Author(s)

Maintainer: Daniel Antal daniel.antal@dataobservatory.eu (ORCID)

Authors:

• Daniel Antal daniel.antal@dataobservatory.eu (ORCID)

Other contributors:

• Marta Kolczynska mkolczynska@gmail.com (ORCID) [contributor]

See Also

Useful links:

• https://retroharmonize.dataobservatory.eu/

• Report bugs at https://github.com/dataobservatory-eu/retroharmonize/issues

Subset surveys from files

Description

Subset surveys from files

Usage

subset_survey_file(
  file_path,
  subset_vars,
  subset_name = "subset",
  id = NULL,
  export_path = NULL
)

Arguments

Subset surveys in memory

Description

Subset surveys in memory

Usage

subset_survey_memory(
  this_survey,
  subset_vars,
  subset_name = "subset",
  export_path = NULL
)

Arguments

Subset and optionally harmonize surveys

Description

Subset one or more surveys by retaining a specified set of variables. Subsetting can be performed either on surveys already loaded in memory or directly from survey files on disk.

If a crosswalk table is supplied, variables are selected based on the variables listed for each survey in the crosswalk, and variable names can optionally be harmonized using 'var_name_target'.

This function replaces the deprecated helpers [subset_waves()] and [subset_save_surveys()].

Usage

subset_surveys(
  survey_list,
  survey_paths = NULL,
  rowid = "rowid",
  subset_name = "subset",
  subset_vars = NULL,
  crosswalk_table = NULL,
  import_path = NULL,
  export_path = NULL
)

subset_waves(waves, subset_vars = NULL)

subset_save_surveys(
  crosswalk_table,
  subset_name = "subset",
  survey_list = NULL,
  subset_vars = NULL,
  survey_paths = NULL,
  import_path = NULL,
  export_path = NULL
)

Arguments

Details

The function supports multiple workflows:

* **In-memory subsetting** using 'survey_list' * **File-based subsetting** using 'survey_paths' or 'import_path' * **Crosswalk-driven subsetting**, where variables are selected per survey using a crosswalk table created by [crosswalk_table_create()]

If 'export_path' is provided, subsetted surveys are written to disk as '.rds' files. Otherwise, subsetted surveys are returned in memory.

Value

Either: * a list of subsetted survey objects (if 'export_path = NULL'), or * a character vector of filenames written to 'export_path'.

See Also

[crosswalk_table_create()], [harmonize_survey_variables()], [read_surveys()]

Examples

examples_dir <- system.file("examples", package = "retroharmonize")
survey_files <- dir(examples_dir, pattern = "\\.rds$")

surveys <- read_surveys(
  file.path(examples_dir, survey_files),
  export_path = NULL
)

subset_surveys(
  survey_list = surveys,
  subset_vars = c("rowid", "isocntry", "qa10_1", "qa14_1"),
  subset_name = "example_subset"
)

Create a survey data frame

Description

Store the data of a survey in a tibble (data frame) with a unique survey identifier, import filename, and optional document object identifier.

Usage

survey(object = data.frame(), id = "survey_id", filename = NULL, doi = NULL)

is.survey(object)

## S3 method for class 'survey'
summary(object, ...)

Arguments

Details

Whilst you can create a survey object with this helper function, it is most likely that you will receive it with an importing function, i.e. read_rds, read_spss read_dta, read_csv or their common wrapper read_survey.

Value

A tibble with id, filename, doi metadata information.

See Also

Other importing functions: is.survey_df()

Examples

example_survey <- survey(
  object = data.frame(
    rowid = 1:6,
    observations = runif(6)
  ),
  id = "example",
  filename = "no_file"
)

Validate harmonize_labels parameter Check if "from", "to", and "numeric_values" are of equal lengths.

Description

Validate harmonize_labels parameter Check if "from", "to", and "numeric_values" are of equal lengths.

Usage

validate_harmonize_labels(harmonize_labels)

Convert labelled missing values to NA

Description

Internal helper used by numeric summary methods to replace SPSS-style missing values with 'NA'.

Usage

vec_convert_na(x)

Arguments

Value

A numeric vector with missing values converted to 'NA'.