traditional2mirt: Convert traditional IRT metric into slope-intercept form used...

Description Usage Arguments Details Value Examples

View source: R/SingleGroup-methods.R

Description

This is a helper function for users who have previously available traditional/classical IRT parameters and want to know the equivalent slope-intercept translation used in mirt. Note that this function assumes that the supplied models are unidimensional by definition (i.e., will have only one slope/discrimination). If there is no supported slope-interecept transformation available then the original vector of parameters will be returned by default.

Usage

1
traditional2mirt(x, cls, ncat)

Arguments

x

a vector of parameters to tranform

cls

the class or itemtype of the supplied model

ncat

the number of categories implied by the IRT model

Details

Supported class transformations for the cls input are:

Rasch, 2PL, 3PL, 3PLu, 4PL

Form must be: (discrimination, difficulty, lower-bound, upper-bound)

graded

Form must be: (discrimination, difficulty 1, difficulty 2, ..., difficulty k-1)

gpcm

Form must be: (discrimination, difficulty 1, difficulty 2, ..., difficulty k-1)

nominal

Form must be: (discrimination 1, discrimination 2, ..., discrimination k, difficulty 1, difficulty 2, ..., difficulty k)

Value

a named vector of slope-intercept parameters (if supported)

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
# classical 3PL model
vec <- c(a=1.5, b=-1, g=.1, u=1)
slopeint <- traditional2mirt(vec, '3PL', ncat=2)
slopeint

# classical graded model (four category)
vec <- c(a=1.5, b1=-1, b2=0, b3=1.5)
slopeint <- traditional2mirt(vec, 'graded', ncat=4)
slopeint

# classical generalize partial credit model (four category)
vec <- c(a=1.5, b1=-1, b2=0, b3=1.5)
slopeint <- traditional2mirt(vec, 'gpcm', ncat=4)
slopeint

# classical nominal model (4 category)
vec <- c(a1=.5, a2 = -1, a3=1, a4=-.5, d1=1, d2=-1, d3=-.5, d4=.5)
slopeint <- traditional2mirt(vec, 'nominal', ncat=4)
slopeint

mirt documentation built on Dec. 11, 2021, 9:20 a.m.