microdadosBrasil

work in progress

NEW:

Censo 2010
Don't use R? See: using the package from Stata and Python

COMMING SOON:

vignettes and Portuguese documentation
RAIS and CAGED (public versions) , PNAD_Continua
Censo 1991, PNADs before 2001
Support for data not fitting into memory.

IN near the future:

variable name harmonization

Description

this package contains functions to read most commonly used Brazilian microdata easily and quickly. Importing these microdata can be tedious. Most data is provided in fixed width files (fwf) with import instructions only for SAS and SPSS. The Data sometimes comes subdivided into serveral files, by state or macro regions. Also, it is common for file and variable names of the same dataset to vary overtime. microdadoBrasil handles all these idiosyncrasies for you. In the background the package is running readr for fwf data and data.table for .csv data. Therefore reading is reasonably fast.

Currently the package includes import functions for:

Source	Dataset	Import_function	Period	Subdataset
IBGE	PNAD	read_PNAD	2001 to 2014	domicilios, pessoas
IBGE	Censo Demográfico	read_CENSO	2000	domicilios, pessoas
IBGE	PME	read_PME	2002.01 to 2015.12	vinculos
IBGE	POF	read_POF	2008	several, see details
INEP	Censo Escolar	read_CensoEscolar	1995 to 2014	escolas, ..., see details
INEP	Censo da Educ. Superior	read_CensoEducacaoSuperior	1995 to 2014	see details
MTE	CAGED	read_CAGED	2009.01 to 2016.05	vinculos
MTE	RAIS	read_RAIS	1998 to 2014	estabelecimentos,vinculos

For the datasets in fwf format, the package includes, internally, a list of import dictionaries. These were constructed with the import_SASdictionary function, which users can use to import dictionaries from datasets not included here. Import dictionaries for the datasets included in the package can be accessed with the get_import_dictionary function.

The package also harmonizes folder names, folder structure and file name that change overtime through a metadata table.It also unifies data that comes subdivides by regional subgroups (UF or região) into a single data object.

Installation

install.packages("devtools")
install.packages("stringi") 
devtools::install_github("lucasmation/microdadosBrasil")
library('microdadosBrasil')

Usage

##############
# Censo Demográfico 2000
download_sourceData("CENSO", 2000, unzip = T)
d <- read_CENSO('domicilios',2000)
d <- read_CENSO('pessoas',2000)

#To import data from a specific path in your machine use:
d <- read_CENSO('domicilios',2000, root_path ="C:/....")
#To restrict the import to only one State use:
d <- read_CENSO('pessoas',2000, UF = "DF")


##############
# PNAD 2002
download_sourceData("PNAD", 2002, unzip = T)
d  <- read_PNAD("domicilios", 2002)
d2 <- read_PNAD("pessoas", 2002)

##############
# Censo Escolar
download_sourceData('CensoEscolar', 2005, unzip=T)
d <- read_CensoEscolar('escola',2005)
d <- read_CensoEscolar('escola',2005,harmonize_varnames=T)

##############
#RAIS
#It will download files for all states and the selected year
download_sourceData("RAIS", i = "2000")
#To import data from all UFs in a single data.frame
d<- read_RAIS('vinculos', i = 2000)
#To import data from a selected group of UFs
d<- read_RAIS('vinculos', i = 2000, UF = c("DF","GO"))

##############
#PME

#It will download files for all months and the selected year:
download_sourceData("PME", i = "2012.01")
#'Period' argument should with quotes and formated as "YYYY.MM"
d <- read_PME("vinculos", "2012.01")

Related efforts

This package is highly influenced by similar efforts, which are great time savers, vastly used and often unrecognized:

Anthony Damico's scripts to read most IBGE surveys. Great if you your data does not fit into memory and you want speed when working with complex survey design data.
Data Zoom by Gustavo Gonzaga, Cláudio Ferraz and Juliano Assunção. Similar ease of use and harmonization of Brazilian microdada for Stata.
dicionariosIBGE, by Alexandre Rademaker. A set of data.frames containing the information from SAS import dictionaries for IBGE datasets.
IPUMS. Harmonization of Census data from several countries, including Brasil. Import functions for R, Stata, SAS and SPSS.

microdadosBrasil differs from those packages in that it:

updates import functions to more recent years
includes non-IBGE data, such as INEP Education Census and MTE RAIS (de-identified)
separates import code from dataset specific metadata, as explained bellow.

How the package works

Traditional Import Workflow

Nowadays packages are normally provided on-line (or in a physical CD for the older IBGE publications) as .zip files with the following structure:

dataset_year.zip

dataset_year
- DICTIONARIES
  - import_dictionary_subdatasetA.SAS
- DATA
  - subdatasetA_state1.txt
  - subdatasetA_state2.txt
  - ...
  - subdatasetA_stateN.txt
- ADITIONAL DOCUMENTATION
  - subdatasetA_variables_and_cathegories_dictionary.xls

Users then normally manually reconstruct the import dictionaries in R by hand. Then, using this dictionary, run the import function, pointing to the DATA folder. Larger datasets (such as CENSUS or RAIS) come subdivided by state (or region), so the function must be repeated for all states. Then if the user needs more than one year of the dataset, the user repeats all the above, adjuting for changes fine and folder names.