Get simulation based cutoff values for G^2^ local dependency test

Based on Chen & Thissen (2007, DOI: 10.2307/1165285), uses package mirt (Chalmers, 2012).

Usage

RIgetResidCorG2(data, iterations = 500, cpu = 4)

Arguments

data

Dataframe with response data

iterations

Number of simulation iterations (needed)

cpu

Number of CPU cores to use (4 is default)

Details

Uses a dataframe with response data to simulate residual correlation values across n simulations based on estimated item & person locations.

Results include mean, max and difference between the mean and max for each iteration. Also, 95th, 99th, 99.5th and 99.9th percentile values are calculated for use with RIresidcorr() as cutoff value, since the max value may be spurious and dependent on number of iterations.

Uses multi-core processing. To find how many cores you have on your computer, use parallel::detectCores(). Remember to keep 1-2 cores free.