In antaldaniel/retroharmonize: Ex Post Survey Data Harmonization

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(retroharmonize)
library(dplyr)

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

Working With a Single Survey

survey_1 <- read_rds(file.path(examples_dir, survey_files[1]))

This function should be renamed and slightly rewritten, it does too many things.

metadata_create(survey_1) %>% head()

Working With Multiple Surveys

survey_paths <- file.path(examples_dir, survey_files)

With smaller data frames representing your surveys, the most efficient way to work with the information is to read them into a list of surveys.

Read the surveys into a list object in the memory:

example_surveys <- read_surveys(survey_paths, .f = "read_rds")

Map the metadata contents of the files:

set.seed(2022)
metadata_create(survey_list = example_surveys) %>%
  sample_n(12)

If you may ran out of memory, you can work with files. The advantage of keeping the surveys in memory is that later it will be much faster to continue working with them, but from the metadata point of view, the returned object is the same either way.

example_metadata <- metadata_create ( survey_paths = survey_paths, .f = "read_rds")

set.seed(2022)
example_metadata %>%
  sample_n(12)

A quick glance at some metadata:

library(dplyr)
subset_example_metadata <- example_metadata %>%
  filter ( grepl("trust", .data$var_label_orig) ) %>%
  filter ( grepl("european_parliament", .data$var_label_orig)) %>%
  select ( all_of(c("filename", "var_label_orig", "var_name_orig", "valid_labels", "na_labels", "class_orig")))

subset_example_metadata

In ZA5913.rds the Trust in European Parliament variable is called qa10_1, in the other surveys it is called qa14_1.

In the first survey, the variable has two values (coded as 1 and 2, and labelled as Tend to trust and Tend not to trust. )

unlist(subset_example_metadata$valid_labels[1])

In the first survey, the variable has two values (coded as 1 and 2, and labelled as Tend to trust and Tend not to trust.) In the second survey, we have three values, and non of them are marked as special, missing values. This is not surprising, because they were not SPSS files. They have related, but not exactly matching classes, too. Therefore, these variables need to be harmonized.

unlist(subset_example_metadata$valid_labels[2])
unlist(subset_example_metadata$na_labels[2])

The metadata created by the metadata_create() and its version for multiple surveys, metadata_create, gives a first overview for the harmonization of concepts, the necessary harmonization of variable names and variable labels. In this case:

Variable name harmonization is required: For a successful join, you must use a common name for qa10_1 and qa14_1, for example, trust_european_parliament, because the variable refers to the same concept.
Variable label harmonization is required: You must make sure that the variable has three categories, even if one category, Declined (to answer) is missing from ZA5913.rds.
Type harmonization is needed: For various statistical procedures you must convert the concatenated contents of qa10_1 and qa14_1 into a numerical variable or factor variable. It is practical to use 1="Tend to Trust", 0="Tend not to trust" for calculating the percentage of people trusting the European Parliament, making sure that Decline will get a NA_real_ value for averaging, or creating a factor variable with three levels, for example trust, not_trust, declined.