Reproducible Survey Data Processing with Step Pipelines

Provides a step-based pipeline for reproducible survey data processing, building on the 'survey' package for complex sampling designs. Supports rotating panels with bootstrap replicate weights, and provides a recipe system for sharing and reproducing data transformation workflows across survey editions.

metasurvey

metasurvey is an R package for processing and analysing complex survey data using metaprogramming and reproducible pipelines. It integrates with the survey package and is designed for complex sampling designs and recurring estimations over time (rotating panels, repeated cross-sections).

If you find this useful, please consider giving us a :star: star on GitHub — it helps others discover the project!

Live services

The full stack is deployed and publicly available:

Key features

• Steps: lazy transformation pipeline (step_compute, step_recode, step_rename, step_remove, step_join) executed via bake_steps().
• Recipes: portable, versioned objects that encapsulate harmonisation pipelines with automatic documentation (doc()) and validation (validate()).
• Workflows: estimation with survey::svymean, svytotal, svyratio and svyby integrated in workflow(), returning a data.table with value, standard error and coefficient of variation.
• Rotating panels: support for RotativePanelSurvey with implantation and follow-ups, and PoolSurvey for combined estimation.
• Replicate weights: bootstrap replicate configuration via add_replicate() for robust variance with survey::svrepdesign.
• Recipe registry: publish, search and discover recipes and workflows through a self-hosted REST API or a local JSON registry.
• Shiny app: interactive recipe and workflow explorer with explore_recipes().
• Self-hosting: deploy the full stack on your infrastructure with Docker Compose or Kubernetes. Publish indicators with full traceability (indicator → workflow → recipe) while keeping microdata private. See vignette("self-hosting").
• STATA transpiler: convert .do files into reproducible Recipe objects.

Works with any household survey

The step pipeline and workflow system are survey-agnostic. The same verbs process Argentina's EPH, Chile's CASEN, Brazil's PNAD-C, the US CPS, Mexico's ENIGH, or DHS data from 90+ countries.

# Same pipeline, different surveys ─────────────────────────

# Argentina (eph)
eph_svy <- Survey$new(
  data = as.data.table(eph::get_microdata(2023, 3)),
  edition = "2023-T3", type = "eph", psu = NULL,
  engine = "data.table", weight = add_weight(quarterly = "PONDERA")
)

# Chile (casen)
casen_svy <- Survey$new(
  data = as.data.table(casen::descargar_casen_github(2017)),
  edition = "2017", type = "casen", psu = "varunit",
  engine = "data.table", weight = add_weight(annual = "expr")
)

# Both use the exact same verbs
process <- function(svy) {
  svy |>
    step_recode(employed, labor_status == 1 ~ 1L, .default = 0L,
                comment = "Binary employment indicator") |>
    bake_steps()
}

See vignette("international-surveys") for reproducible examples with all seven surveys.

Installation

Install the stable version from CRAN:

install.packages("metasurvey")

Or the development version from GitHub:

# install.packages("devtools")
devtools::install_github("metasurveyR/metasurvey")

Quick example

library(metasurvey)

# Create a survey with sample data
data(api, package = "survey")

svy <- Survey$new(
  data    = apistrat,
  edition = "2000",
  type    = "api",
  psu     = NULL,
  engine  = "data.table",
  weight  = add_weight(annual = "pw")
)

# Lazy transformations
svy <- step_compute(svy, growth = api00 - api99, comment = "API growth")
svy <- step_recode(svy, school_level,
  stype == "E" ~ "Elementary",
  stype == "M" ~ "Middle",
  stype == "H" ~ "High",
  .default = NA_character_
)
svy <- bake_steps(svy)

# Estimation
workflow(
  list(svy),
  survey::svymean(~growth, na.rm = TRUE),
  estimation_type = "annual"
)

Full example: ECH panel with bootstrap replicate weights

This example uses the rotating panel from Uruguay's Encuesta Continua de Hogares (ECH) with bootstrap replicate weights. First, download the example data:

download_example_ech <- function() {
  zip_url <- "https://informe-tfg.s3.us-east-2.amazonaws.com/example-data.zip"
  dest_zip <- "example-data.zip"
  temp_dir <- tempfile("example-data")
  download.file(zip_url, destfile = dest_zip, mode = "wb")
  dir.create(temp_dir)
  unzip(dest_zip, exdir = temp_dir)
  target_dir <- "example-data"
  dir.create(target_dir, recursive = TRUE, showWarnings = FALSE)
  file.rename(
    list.files(file.path(temp_dir, "example-data"), full.names = TRUE),
    file.path(target_dir, basename(list.files(file.path(temp_dir, "example-data"))))
  )
  unlink(dest_zip)
  unlink(temp_dir, recursive = TRUE)
}
download_example_ech()

With the data downloaded:

library(metasurvey)
library(magrittr)

path_dir <- file.path("example-data", "ech", "ech_2023")

ech_2023 <- load_panel_survey(
  path_implantation = file.path(path_dir, "ECH_implantacion_2023.csv"),
  path_follow_up = file.path(path_dir, "seguimiento"),
  svy_type = "ECH_2023",
  svy_weight_implantation = add_weight(annual = "W_ANO"),
  svy_weight_follow_up = add_weight(
    monthly = add_replicate(
      "W",
      replicate_path = file.path(
        path_dir,
        c(
          "Pesos replicados Bootstrap mensuales enero_junio 2023",
          "Pesos replicados Bootstrap mensuales julio_diciembre 2023"
        ),
        c(
          "Pesos replicados mensuales enero_junio 2023",
          "Pesos replicados mensuales Julio_diciembre 2023"
        )
      ),
      replicate_id = c("ID" = "ID"),
      replicate_pattern = "wr[0-9]+",
      replicate_type = "bootstrap"
    )
  )
)

# Build labour market indicators
ech_2023 <- ech_2023 %>%
  step_recode("pea", POBPCOAC %in% 2:5 ~ 1, .default = 0,
              comment = "EAP", .level = "follow_up") %>%
  step_recode("pet", e27 >= 14 ~ 1, .default = 0,
              comment = "WAP", .level = "follow_up") %>%
  step_recode("po", POBPCOAC == 2 ~ 1, .default = 0,
              comment = "Employed", .level = "follow_up") %>%
  step_recode("pd", POBPCOAC %in% 3:5 ~ 1, .default = 0,
              comment = "Unemployed", .level = "follow_up")

ech_2023_bake <- bake_steps(ech_2023)

# Quarterly rates: activity, employment and unemployment
workflow_result <- workflow(
  survey = extract_surveys(ech_2023_bake, quarterly = 1:4),
  survey::svyratio(~pea, denominator = ~pet),
  survey::svyratio(~po, denominator = ~pet),
  survey::svyratio(~pd, denominator = ~pea),
  estimation_type = "quarterly:monthly",
  rho = 0.5,
  R = 5 / 6
)

workflow_result

This pipeline loads a rotating panel with bootstrap replicate weights, builds binary labour market indicators (EAP, WAP, employed, unemployed), and estimates activity, employment and unemployment rates by quarter with robust variance.

STATA transpiler

Many research groups maintain decades of STATA .do files that process household survey microdata. The metasurvey transpiler converts these scripts into reproducible Recipe objects.

library(metasurvey)

# Transpile a .do file to metasurvey steps
result <- transpile_stata("demographics.do")
result$steps[1:3]
#> [1] "step_rename(svy, hh_id = \"id\", person_id = \"nper\")"
#> [2] "step_compute(svy, weight_yr = pesoano)"
#> [3] "step_compute(svy, sex = e26)"

# Transpile an entire year directory into separate recipes
recipes <- transpile_stata_module(
  year_dir = "do_files/2022",
  year = 2022,
  user = "research_team",
  output_dir = "recipes/"
)

# Check coverage before migrating
transpile_coverage("do_files/")

Supported STATA patterns: gen/replace chains, recode, egen with by-groups, foreach/forvalues loops, mvencode, destring, rename, drop/keep, variable and value labels, inrange/inlist expressions, and variable ranges.

See vignette("stata-transpiler") for the full reference.

Documentation

• Getting started
• International survey compatibility
• Recipes
• Workflows and estimation
• Complex designs
• Rotating panels
• Case study: ECH
• STATA transpiler
• ECH demographics recipe
• Interactive explorer
• API and database
• Self-hosting guide

Related work

metasurvey is not a wrapper around survey. It adds a reproducibility layer (steps, recipes, workflows) that is survey-agnostic: the same pipeline processes ECH, EPH, CASEN, PNAD-C, CPS, ENIGH, or DHS data without survey-specific code.

Citation

To cite metasurvey in publications use:

citation("metasurvey")

Loprete M, da Silva N, Machado F (2026). metasurvey: Reproducible Survey Data Processing with Step Pipelines. R package, https://CRAN.R-project.org/package=metasurvey.

Contributing

Please see CONTRIBUTING.md for guidelines on how to contribute to metasurvey.

Code of Conduct

Please note that the metasurvey project is released with a Contributor Code of Conduct. By contributing to this project you agree to abide by its terms.