betaFeedback

WorksPhysics and Astronomy

Microstate Response Model: A Unified IRT Framework from Latent Nominal States to a Theory of Tests

Abstract

Several features outside of the conventional assumptions in traditional item response theory (IRT) models have been revealed to be related, yet no unifying framework explains the features simultaneously. They can be unified by treating each cause as a distinct reason, and then regarding every reason as a choice in a latent multiple-choice item. Here, we propose the Microstate Response Model (MiRM), which decomposes each response category (macrostate) into multiple latent reasons (microstates) using Boltzmann weights. The consequences of this simple framework are substantial; for example, it reads the lower and upper asymptotes of item characteristic curves (ICCs) as additional states, and ICC asymmetry and item complexity as multiple states. The model unifies many classical models from the dichotomous logistic family to mixture models. Moreover, the algebra of the model provides the intuition that items and tests are the same object. Two consequences naturally come out: partial credits and test scores are unified, and a new test-level monotonicity condition is proposed. We examine item characteristics including nonmonotonicity on PISA 2015, and show that MiRM provides a deeper understanding of items while achieving the strongest overall performance across fit, parsimony, and prediction.

Paper details

vlnc:W_s27kvgrw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
arXiv nonexclusive license
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…