Simulated and synthetic (simsyn) data

The IRW includes a data family of simulated and synthetic item response datasets, collectively referred to as simsyn. These data follow the same IRW data standard as the core repository and can be accessed using the same irw package functions, with source = "sim".

• Simulated data are generated from a statistical model (e.g., a Rasch or 2PL model) with known parameter values. They are useful for methods development, power analysis, and verifying that estimation procedures recover the true parameters.

• Synthetic data are constructed to approximate the properties of a specific real dataset — preserving features like the number of respondents, items, and response distributions — while protecting privacy or enabling open sharing where the original cannot be released.

Simsyn is the one family defined by where the data come from rather than by their shape. A simulated table can be dichotomous, polytomous or nominal; it belongs here, not in the core or nominal family, because no real person produced the responses.

Accessing the simsyn data

Documentation for the available tables can be found here.

R

library(irw)

tables <- irw_list_tables(source = "sim")     # List available simsyn tables
df <- irw_fetch(tables$name[1], source = "sim") # Fetch a simsyn table

Python

import irw

tables = irw.list_tables(source="sim")        # List available simsyn tables
df = irw.fetch(tables["name"].iloc[0], source="sim")  # Fetch a simsyn table

What the family holds

Each table links to its own page, with citation, licence and fetch snippets.

Table	Respondents	Items	Responses	Categories	Licence	Source
aberrance_pk_sim	1,000	40	40,000	2	CC0 1.0	link
aberrance_speeded_sim	1,000	40	40,000	2	CC0 1.0	link
dtmr_sim	990	27	26,730	2	MIT	link
estcrm	1,000	5	5,000	0	CC BY 4.0	link
gilbert_meta_3	5,314	17	90,338	2	CC BY 4.0	link
heller2013_probability	345	12	4,140	2	CC BY 4.0	link
hmcdm_learning_sim	1,000	50	50,000	2	CC0 1.0	link
hrm_rater_sim	1,000	6	18,000	4	CC0 1.0	link
ising_network_sim	1,000	20	20,000	2	CC0 1.0	link
lnirt_sim	1,000	20	20,000	2	CC0 1.0	link
mirt_ggum	1,000	10	10,000	3	CC BY 4.0	link
mirt_mixed	2,000	6	12,000	3	CC BY 4.0	link
mirt_nonlin	2,000	6	12,000	2	CC BY 4.0	link
mudfoldsim	1,000	10	10,000	2	CC0 1.0	link
nbashots_sim	400	5	136,368	2	CC BY 4.0	link
raschmix_rost2_sim	1,800	10	18,000	2	CC0 1.0	link
tirt_sim_trt	500	14	7,000	4	CC0 1.0	link

Showing 17 of 17 tables. Fetch any of them with irw_fetch("<table>", source = "sim") or irw.fetch("<table>", source="sim").

Use cases

Simsyn data are particularly useful when you want to:

• Benchmark estimation methods against data with known ground-truth parameters
• Develop or test new psychometric approaches without relying on real participants
• Reproduce analyses from published work that used proprietary data, via a synthetic stand-in
• Teach IRT concepts using data that behave predictably under a given model

For realistic simulation of item difficulty parameters informed by the empirical IRW distribution, see the Simulating Item Difficulties vignette.