baggedParams: Function to calculate bagged IRT parameters from a score...

In talentlens/talentlens: R package for team TalentLens

Description

Performs IRT fitting algorithm repeatedly (R times) on a random sample of the score matrix (with replacement), according to a bootstrap procedure. The results are then aggregated (bagging) and a list of parameter and error estimates is returned.

Usage

baggedParams(score_mat, R, model = c("2pl", "rasch"),
central.fun = mean, spread.fun = sd)

Arguments

score_mat	(nxk) numeric matrix rows corresponding to participants and columns corresponding to items. NOTE: columns need to be named after item ids.
R	Number of iterations for the bootstrap loop. Higher R increases the time the function needs for completion, but also gives better estimates of bagged parameters and bootstrapped errors. Default: 199 (on a Dell laptop, this takes about 1 min to run on a dataset with 99 rows).
model	The IRT model to use. This function currently supports rasch and 2pl models.
central.fun	The function used for aggregating results and find the central values. Default: mean.
spread.fun	The function used for estimating parameter errors. Default: sd.

Value

params	A (kx2) matrix of IRT parameters
error	A (kx2) matrix of estimation errors

Author(s)

Morgan Strom

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