A Comparison of Linking and Concurrent Calibration Under Item Response Theory

Kim, Seock-Ho; Cohen, Allan S.

doi:10.1177/01466216980222003

Cited by 99 publications

(98 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…This finding is consistent with the results of Gök (2012), which showed that sample size had no positive effect on methods, while test length had a positive effect. In addition, studies have shown that equating performance is in accordance with equating steps and therefore, a one-step process is better than a two-step process (Chu, 2002;Hanson & Beguin, 1999b;Kim & Cohen, 1998). Consistent with other studies, the present study found the highest bias value in the IRM-SC.…”

Section: Discussionsupporting

confidence: 90%

Effect of Differential Item Functioning on Test Equating

2015

EDUC SCI-THEOR PRACT

View full text Add to dashboard Cite

This study examines the effect of differential item functioning (DIF) items on test equating through multilevel item response models (MIRMs) and traditional IRMs. The performances of three different equating models were investigated under 24 different simulation conditions, and the variables whose effects were examined included sample size, test length, DIF magnitude, and the test type. The MIRMs, in which the DIF factors were added as parameters, were compared with the Stocking-Lord (SL) method (one of the IRM-based calibration methods) and concurrent calibration method. According to the results, differences were found in the performances of the methods under the analyzed conditions. More specifically, the MIRMs were able to identify the DIF items, carry out the equating processes, and eliminate the biases caused by DIF in only one analysis. However, this does not indicate that using MIRMs is the best approach since the increase in sample size and test length generally had a positive effect on IRM-based equating, whereas MIRMs were less affected by these two conditions. Considering the IRM-based methods, it was found that separate calibration methods were more affected by the presence of DIF items compared to concurrent calibration. Moreover, this effect becomes most significant when DIF items are in common test and the magnitude of DIF is C.

show abstract

Section: Discussionsupporting

confidence: 90%

Effect of Differential Item Functioning on Test Equating

2015

EDUC SCI-THEOR PRACT

View full text Add to dashboard Cite

show abstract

“…Two studies indicated that concurrent calibration as is implemented in BILOG-MG performed equally well in comparison with traditional methods in vertical equating (Béguin & Hanson, 2001;Kim & Cohen, 1998). Because in our study the number of anchor items was large and because these anchor items did form a representative sample of the test, technical issues of vertical equating should not be a problem here.…”

Section: Discussionmentioning

confidence: 72%

“…A formal exposure can be found in Bock and Zimowski (1995) and Mislevy (1987). Kim and Cohen (1998) compared concurrent calibration methods with the approach based on separate calibrations using equating coefficients from the 1042 EDUCATIONAL AND PSYCHOLOGICAL MEASUREMENT characteristic curve method. They used a simulated data set consisting of two groups differing in ability level.…”

Section: Developmental Score Scales and Vertical Equatingmentioning

confidence: 99%

Investigating the Dimensions of Spelling Ability

Notenboom¹,

Reitsma

2003

Educational and Psychological Measurement

View full text Add to dashboard Cite

In this study, the latent structure of a spelling achievement test for elementary school grades was investigated. Factor analyses revealed that for Grades 2 to 6, the test was unidimensional, whereas for Grade 1, two factors were found: a phonological factor and an orthographic factor. Because of the anchor-test design, vertical equatingwas required to fit a two-parameter, unidimensional item response theory model. Concurrent calibration of the parameters, implemented in BILOG–Multiple Groups (MG), was used as the linking procedure. An evaluation of the model showed that invariant parameter estimates and ability scores were obtained. The transition from an alphabetic to an orthographic strategy occurs in the first grade. Learning to spell throughout elementary education mainly consists of learning to apply orthographic knowledge.

show abstract

“…Using the marginal maximum likelihood estimation method, Kim and Cohen (1998) examined the separate calibration method using BILOG (Mislevy & Bock, 1982) with the Stocking and Lord method (Stocking & Lord, 1983) and the concurrent calibration method using MULTILOG (Thissen, 1991). They concluded that the two methods provided similar results except when the number of common items was small (e.g., 5 out of 50), where separate calibration provided more accurate results.…”

Section: Concurrent Vs Separate Calibrations In Uirtmentioning

confidence: 99%

“…However, Hanson and Beguin (2002) pointed out that the differences between concurrent and separate calibration results in the case of non-equivalent groups in the Kim and Cohen (1998) study were confounded with the different computer programs:…”

Section: Concurrent Vs Separate Calibrations In Uirtmentioning

confidence: 99%

Exploring the Full-Information Bifactor Model in Vertical Scaling With Construct Shift

Lissitz

2012

Applied Psychological Measurement

View full text Add to dashboard Cite

To address the lack of attention to construct shift in IRT vertical scaling, a bifactor model is proposed to estimate the common dimension for all grades and the grade-specific dimensions. The bifactor model estimation accuracy is evaluated through a simulation study with manipulated factors of percent of common items, sample size, and degree of construct shift. In addition, the unidimensional IRT (UIRT) estimation model that ignores construct shift is examined to represent the current practice for IRT vertical scaling; comparisons on parameter estimation accuracy of the bifactor and UIRT models are discussed.The major findings of the simulation study are (1) bifactor models are well recovered overall, even though item discrimination parameters are underestimated to a small degree; (2) item discrimination parameter estimates are overestimated in UIRT models due to the effect of construct shift; (3) person parameters of UIRT models are less accurately estimated than that of bifactor models, and the accuracy decreases as the degree of construct shift increases; (4) group mean parameter estimates of UIRT models are less accurate than that of bifactor models, and a large effect due to construct shift is found for the group mean parameter estimates of UIRT models.The real data analysis provides an illustration of how bifactor models can be applied to a problem involving for vertical scaling with construct shift. General procedures for testing practice are also discussed.

show abstract

A Comparison of Linking and Concurrent Calibration Under Item Response Theory

Cited by 99 publications

References 16 publications

Effect of Differential Item Functioning on Test Equating

Effect of Differential Item Functioning on Test Equating

Investigating the Dimensions of Spelling Ability

Exploring the Full-Information Bifactor Model in Vertical Scaling With Construct Shift

Contact Info

Product

Resources

About