A Datasets used in the book

Every example loads its data in one chunk, labelled data-<chapter>-<n>, at the start of the example. The table lists what each chunk defines. Real data come with mirt; simulated data are generated in the chunk itself, so the true parameters are known and the examples can check estimates against them.

Table A.1: Data chunks in the book and the objects each defines.
Chunk Data Kind Defines
data-measurement-1 SAT12: 600 students, 32 five-option science items Real raw, key
data-measurement-2 8-item wellbeing scale, 0 to 3, two negatively worded items Simulated resp
data-irt-models-1 LSAT7: 1,000 test takers, 5 items Real resp
data-irt-models-2 20-item, four-option driving theory test with guessing Simulated resp, true
data-estimation-1 LSAT6: 1,000 test takers, 5 items Real resp
data-estimation-2 SAT12, scored with the corrected key Real resp, key
data-dimensionality-1 24-item mathematics test with calculation and reading dimensions Simulated resp, a, d, sigma
data-dimensionality-2 20-item reading test with a four-item passage Simulated resp
data-fit-1 SAT12, scored with the corrected key Real resp, key
data-fit-2 30-item quiz with random responders and pre-knowledge Simulated resp, type, a, b
data-information-1 SAT12, scored with the corrected key Real resp, key
data-information-2 Item parameters of a 90-item nursing pool Simulated pool
data-scoring-1 LSAT7 Real resp
data-scoring-2 40-item mathematics test, true abilities known Simulated resp, true_theta
data-polytomous-1 Science: 392 respondents, 4 attitude items scored 1 to 4 Real resp
data-polytomous-2 6-item job satisfaction scale, 0 to 4, one item with disordered steps Simulated resp
data-dif-1 25-item reading test taken on paper or tablet Simulated resp, group
data-dif-2 20-item chemistry quiz by gender Simulated resp, group
data-linking-1 Two 30-item placement forms with 10 anchors (NEAT design) Simulated old, new, anchors, pars
data-linking-2 The same two forms stacked for concurrent calibration Simulated resp, form
data-banks-cat-1 Parameters of a 200-item vocabulary bank in four areas Simulated bank
data-banks-cat-2 300 simulated test takers for the CAT Simulated true_theta
data-reporting-1 36-item Grade 6 science test Simulated resp
data-reporting-2 Science Real resp
data-ml-lens-1 20-item test with known abilities and item parameters Simulated resp, theta, a, b
data-ml-lens-2 40-item placement test and a pass/fail course result Simulated resp, ability, passed
data-item-text-1 70 science classification items with text features and the responses of 1,000 students Simulated sci, resp, rasch
data-aig-1 Three item models for Grade 7 items: Average speed, a fraction of a number and the mean Simulated speed, frac, avg, fams
data-aig-2 A field test of the 36 generated items, answered by 800 students Simulated resp, rasch, items
data-ai-validity-1 Machine and human scores of two writing prompts, by device and testing window Simulated writing
data-ai-validity-2 Field-test difficulties and features of 45 items from three item models Simulated items
data-auto-scoring-1 Short answers to a science question, scored 0 to 3 by one or two raters Simulated bread
data-auto-scoring-2 Four answers with a known correct score Simulated probe
data-auto-scoring-3 A test with 10 multiple-choice items and three written answers scored by two raters and an engine Simulated written, mc
data-llms-1 Science Real resp
data-llms-2 A two-cell blueprint for a numeracy quiz Simulated cells
data-llms-3 hexaco_real and hexaco_llm: 40 Honesty-Humility items, real Canadian respondents and four language models (4 attempts each) Real and generated real, synthetic
data-knowledge-graphs-1 fraction_subtraction: 536 students, 20 fraction subtraction items, with its 8-attribute Q-matrix as a knowledge graph Real resp, kg
data-knowledge-tracing-1 cognitive_tutor: Cognitive Tutor practice logs, 587 students, 16,857 attempts on 12 skills Real logs
data-knowledge-tracing-2 fractions_logs and fractions_kg: Practice logs of 400 students on the fractions graph Simulated practice, kg

The book deliberately does not analyse the mirtverse lab datasets (certexam, numeracy, linking, catbank and openended), because the labs ask students to discover problems hidden in them. The same goes for the AI Hub datasets itemtext and essays, and the knowledge graph example fractions. They are described in the teaching labs article and on their help pages, for example ?certexam.

Other datasets for practice

mirt ships more datasets that work well for exercises:

  • LSAT6 and LSAT7: Short dichotomous tests (expand them with expand.table()).
  • deAyala: Five dichotomous mathematics items from de Ayala’s textbook, as a frequency table.
  • SLF: Five dichotomous statements about social life feelings, answered by 1,490 respondents.
  • Bock1997: A three-item polytomous test, as a frequency table.
  • ASVAB: Four arithmetic reasoning items with response counts for four groups defined by sex and race, a ready-made exercise for multiple-group models and DIF.

Use ?<name> for the details of each.