EXPLORING THE APPLICABILITY OF 4-PARAMETER LOGISTIC ITEM RESPONSE THEORY MODEL IN THE
CALIBRATION OF COMPUTER-BASED MATHEMATICS ACHIEVEMENT TEST
Abstract
This study explored the applicability of 4-parameter logistic Item
Response Theory (IRT) model in the calibration of senior secondary
school (SSS) Computer-Based Mathematics Achievement Test (CBMAT)
in Lagos State, Nigeria. A causal-comparative design was adopted while
Lagos Education District 1 was used with eight schools that had
functional computer laboratories where 874 SSS 2 students were
randomly selected. The CBMAT had IRT empirical reliability value of
0.89. Two research questions were raised and data was analysed using
Stout's Test of Essential Dimensionality and 1-, 2-, 3- and 4PL models of
Multidimensional IRT package of R programming. Results revealed that
the CBMAT instrument was unidimensional and fitted the 4PL model.
The popularly used models for calibration by researchers for so many
years are the 1, 2 and 3PL models. This is due to the complexity that was
attached to estimating parameters of the 4PL model. However, software
has been made available to tackle this issue. It was therefore inferred
from the findings that CBMAT revealed only a dominant mathematics
ability trait. Based on the findings, it was recommended that suitable
calibration procedures for the right choice of appropriate model should
be imbibed by test developers, pyschometricians and researchers on any
form of scale that is meant to measure students' learning outcomes.
Keywords: 4-parameter logistic IRT model, calibration, computer-based mathematics
achievement test, unidimensional.
Omonike R. Lawal & J. Gbenga Adewale
EXPLORING THE APPLICABILITY OF 4-PARAMETER LOGISTIC ITEM RESPONSE THEORY MODEL IN THE CALIBRATION OF COMPUTER-BASED MATHEMATICS ACHIEVEMENT TEST
Abstract This study explored the applicability of 4-parameter logistic Item Response Theory (IRT) model in the calibration of senior secondary school (SSS) Computer-Based Mathematics Achievement Test (CBMAT) in Lagos State, Nigeria. A causal-comparative design was adopted while Lagos Education District 1 was used with eight schools that had functional computer laboratories where 874 SSS 2 students were randomly selected. The CBMAT had IRT empirical reliability value of 0.89. Two research questions were raised and data was analysed using Stout's Test of Essential Dimensionality and 1-, 2-, 3- and 4PL models of Multidimensional IRT package of R programming. Results revealed that the CBMAT instrument was unidimensional and fitted the 4PL model. The popularly used models for calibration by researchers for so many years are the 1, 2 and 3PL models. This is due to the complexity that was attached to estimating parameters of the 4PL model. However, software has been made available to tackle this issue. It was therefore inferred from the findings that CBMAT revealed only a dominant mathematics ability trait. Based on the findings, it was recommended that suitable calibration procedures for the right choice of appropriate model should be imbibed by test developers, pyschometricians and researchers on any form of scale that is meant to measure students' learning outcomes. Keywords: 4-parameter logistic IRT model, calibration, computer-based mathematics achievement test, unidimensional.