Explore Available Datasets
Before fetching data, you can explore the available datasets and their metadata.
List all datasets and their metadata:
irw_list_tables()
irw_info()
irw_info("4thgrade_math_sirt")Find Datasets by Criteria
Filter datasets based on key attributes.
Example: Filter by respondent count and density range:
irw_filter(n_participants = c(100, 1000), density = c(0.1, 0.5))Find datasets with rater-based scoring:
irw_filter(var = "rt")Combine multiple filters:
irw_filter(n_participants = c(100, 10000), var = "cov_", density = c(0.1, 0.8))Fetch Datasets
Retrieve datasets directly into R for analysis.
Fetch a single dataset:
Fetch multiple datasets:
datasets <- irw_fetch(c("fractals_rating", "spelling2pronounce_edwards2023"))
print(names(datasets))Fetch datasets that match specific filters:
matching_tables <- irw_filter(var = "rater", density = NULL)
metadata <- irw_metadata()
metadata[metadata$table %in% matching_tables, ]
matching_tables <- irw_filter(var = "rater", density = NULL, responses_per_item = c(0, 1000))
datasets <- irw_fetch(matching_tables)
print(names(datasets))Pin a Version for Reproducibility
IRW grows and gets corrected: tables are added, and errors found in published tables are fixed. A script that fetches by name alone will therefore not necessarily return the same data next year. Pin the corpus to an IRW version and it will.
irw_use_version(332) # every fetch below reads v332
swmd_mokken <- irw_fetch("swmd_mokken") # v332's copy, today and next yearThe IRW version number names one exact combination of all eleven
Redivis datasets that make up IRW, so it is the thing to cite in a
paper. Call irw_use_version() with no argument at the start
of a project to pin today’s version and print the number to record:
irw_use_version() # "Reading IRW v332 ... 11 dataset(s) pinned."
irw_get_version() # what the session is reading, dataset by dataset
irw_reset_version() # back to the current releaseThe pin covers metadata and item text as well as response data, and holds for the R session. A table added to IRW after the pinned version is an error rather than a quiet fetch from the current release — mixing versions would defeat the point.
You can also look a version up without reading it:
irw_version() # the newest version and its eleven Redivis tags
irw_version(version = 332) # exactly what v332 held
irw_version("2026-08-01") # what was live on a dateCite the number rather than the date. Redivis overwrote its own release timestamps for the older warehouse shards during a platform migration, so dates before 21 July 2026 resolve only approximately and say so; version numbers are always exact.
Download Datasets
Save datasets locally.
Save as CSV:
irw_download("gilbert_meta_48")Save dataset metadata as a BibTeX citation:
irw_save_bibtex("gilbert_meta_48")