findAnchors: Function to suggest optimal anchor items from a score matrix
In talentlens/talentlens: R package for team TalentLens

Description Usage Arguments Value Author(s)

This function selects optimal anchor items in two steps: first, based on CTT analyses (proportion of participants answering the item correctly and item-total biserial correlation) and second, based on a chi-square test of item fit under the chosen IRT model.

1
2
3

findAnchors(score_matrix, model = c("rasch", "2pl"),
bagging = FALSE, R = NA,
min_cor = 0.2, min_p = 0.1, max_p = 0.9, verbose = FALSE)

`score_matrix`	A numeric matrix where each row represent the scores of a participant, and each row represent a item. NOTE the columns need to be named with item ids!
`model`	The IRT model to use for parameter fitting. Either "rasch" (for smaller sample sizes) or "2pl" (for larger sample sizes)
`bagging`	If TRUE, the parameters will be estimated using bagging (see package baggedParams). If FALSE, parameters will be estimated using one single run of the item fitting algorithm.
`R`	For bagging, this is the number of bootstrap iterations.
`min_cor`	The minimum allowed item-total biserial correlation. Default: 0.2
`min_p`	The minimum allowed proportion of participants answering a given item correctly. Default: 0.1
`max_p`	The maximum allowed proportion of participants answering a given item correctly. Default: 0.9
`verbose`	If TRUE, information about the item selection will be printed to the console.

A two-column matrix of item parameters for the suggested anchor items. The rows are named after the item ids, and the columns are named "b" (difficulty parameter) and "a" (discrimination parameter).

Morgan Strom

talentlens/talentlens documentation built on May 31, 2019, 2:52 a.m.