The InterModel Vigorish (IMV) and IRT

Getting started

The IMV

Key details regarding the use of the IMV when using item response theory (IRT) models to analyze dichotomous item repsonses are discussed below. For a generic introduction to the IMV concept and for illustrations of how it can be used more generally to model binary outcomes, see here.

The irtimv package

The irtimv package gets used at various points and will need to be installed to replicate results.

devtools::install_github("ben-domingue/irtimv/irtimv-helper", ref="main")

A basic example

We offer a few basic examples of how to compute the IMV. The imv function here can be used to compute the IMV with either simulated or empirical data.

Reproducing results from the paper

The simulations

Simulation code is here. Code for simulation studies is largely self-contained (outside of usage of irtimv package). Figures from the main text are created in the following files

• f1: misfit.R
• f2: misfit.R
• f3: altmetric.R

Figures from the SI:

• f1: pvalue.R
• f2: misfit2.R
• f3: misfit_bysumscore.R
• f4: theta.R
• f5: prior_discrimination.R
• f6: prior.R
• f7: Nsim4.R
• f8: Nsim3.R
• f9: fuzzy.R
• f10: altmetric.R
• f12: multidimensional.R

Empirical examples

Code supporting work with empirical data is here. This paper uses data from the Item Response Warehouse (IRW). Figures from the main text are produced via:

• 02_complexity.R computes unidimensional results.
• 03_mirt.R comptues multidimensional results.
• 04_figure.R produces figure 3.