tracedif: Difference in tracelines or TCC

In falkcarl/mpirt: Semi- and non-parametric item response theory

Difference in tracelines or TCC

Description

Difference in tracelines or TCC

Usage

tracedif(
  mod1,
  mod2,
  j1,
  j2,
  method = c("RIMSD", "ISD", "IAD"),
  scale = TRUE,
  scaletest = TRUE,
  theta = seq(-5, 5, 0.1),
  w = dnorm(theta)/sum(dnorm(theta))
)

Arguments

mod1	Fitted mxModel, e.g., from fitMP.
mod2	Fitted mxModel, e.g., from fitMP.
j1	Item index (or vector of indices in the case of TCC) for the first model.
j2	Item index (or vector of indices in the case of TCC) for the second model.
method	Which type of discrepancy to compute? RIMSD, ISD, or IAD.
scale	logical value to determine whether to scale each expected score value to 0-1. See TCC.
scaletest	logical value to determine whether to scale the overall test by dividing by the number of items. See TCC.
theta	Grid over which to compute the traceline or TCC.
w	Vector of weights; usually these should sum to 1.

Details

This is a convenience function which allows computation of RIMSD, ISD, or IAD directly from two fitted models and will do so on either single items from each or for TCCs if a vector of items from each is selected. Note that RIMSD is usually scaled so that the resulting value is on a metric between 0 and 1. Variants of ISD and IAD (wABC an sDTF) in previous papers were dependent on the number of score categories. However, Chalmers et al. mentions a variant that can serve as an effect size if further scaled.

falkcarl/mpirt documentation built on July 11, 2024, 12:09 a.m.