Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data

Letty Koopman

doi:10.1007/978-3-031-27781-8_20

Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data

Research output: Chapter in Book/Report/Conference proceeding › Chapter › Academic › peer-review

Abstract

Investigating model fit is essential for determining measurement properties of tests and questionnaires. Mokken scale analysis (MSA) consists of a selection of methods to investigate the fit of nonparametric item response theory models. Existing MSA methods assume a simple random sample, which is violated in two-level test data (i.e., test data of clustered respondents). This chapter discusses the methods manifest monotonicity, conditional association, and manifest invariant item ordering, and investigates the effect of within-group dependency on the point estimate and variability of their statistics. Results showed that fit statistics may be safely used in the presence of within-group dependency, giving appropriate results for sets of items that either did or did not violate assumptions. Implications for practice are discussed.

Original language	English
Title of host publication	Quantitative Psychology
Subtitle of host publication	The 87th Annual Meeting of the Psychometric Society, Bologna, Italy, 2022
Editors	Marie Wiberg, Dylan Molenaar, Jorge González, Jee-Seon Kim, Heungsun Hwang
Publisher	Springer
Pages	221-232
Number of pages	12
ISBN (Electronic)	978-3-031-27781-8
ISBN (Print)	978-3-031-27780-1, 978-3-031-27783-2
DOIs	https://doi.org/10.1007/978-3-031-27781-8_20
Publication status	Published - 2023
Externally published	Yes

Publication series

Name	Springer Proceedings in Mathematics & Statistics
Publisher	Springer
Volume	422
ISSN (Print)	2194-1009
ISSN (Electronic)	2194-1017

Access to Document

10.1007/978-3-031-27781-8_20

Handle.net

Persistent link

Cite this

Koopman, L. (2023). Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data. In M. Wiberg, D. Molenaar, J. González, J.-S. Kim, & H. Hwang (Eds.), Quantitative Psychology: The 87th Annual Meeting of the Psychometric Society, Bologna, Italy, 2022 (pp. 221-232). (Springer Proceedings in Mathematics & Statistics; Vol. 422). Springer. https://doi.org/10.1007/978-3-031-27781-8_20

Koopman, Letty. / Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data. Quantitative Psychology: The 87th Annual Meeting of the Psychometric Society, Bologna, Italy, 2022. editor / Marie Wiberg ; Dylan Molenaar ; Jorge González ; Jee-Seon Kim ; Heungsun Hwang. Springer, 2023. pp. 221-232 (Springer Proceedings in Mathematics & Statistics).

@inbook{50e30d9798cd48cc90954206bdad45c1,

title = "Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data",

abstract = "Investigating model fit is essential for determining measurement properties of tests and questionnaires. Mokken scale analysis (MSA) consists of a selection of methods to investigate the fit of nonparametric item response theory models. Existing MSA methods assume a simple random sample, which is violated in two-level test data (i.e., test data of clustered respondents). This chapter discusses the methods manifest monotonicity, conditional association, and manifest invariant item ordering, and investigates the effect of within-group dependency on the point estimate and variability of their statistics. Results showed that fit statistics may be safely used in the presence of within-group dependency, giving appropriate results for sets of items that either did or did not violate assumptions. Implications for practice are discussed.",

author = "Letty Koopman",

year = "2023",

doi = "10.1007/978-3-031-27781-8_20",

language = "English",

isbn = "978-3-031-27780-1",

series = "Springer Proceedings in Mathematics & Statistics",

publisher = "Springer",

pages = "221--232",

editor = "Marie Wiberg and Dylan Molenaar and Jorge Gonz{\'a}lez and Jee-Seon Kim and Heungsun Hwang",

booktitle = "Quantitative Psychology",

}

Koopman, L 2023, Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data. in M Wiberg, D Molenaar, J González, J-S Kim & H Hwang (eds), Quantitative Psychology: The 87th Annual Meeting of the Psychometric Society, Bologna, Italy, 2022. Springer Proceedings in Mathematics & Statistics, vol. 422, Springer, pp. 221-232. https://doi.org/10.1007/978-3-031-27781-8_20

Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data. / Koopman, Letty.
Quantitative Psychology: The 87th Annual Meeting of the Psychometric Society, Bologna, Italy, 2022. ed. / Marie Wiberg; Dylan Molenaar; Jorge González; Jee-Seon Kim; Heungsun Hwang. Springer, 2023. p. 221-232 (Springer Proceedings in Mathematics & Statistics; Vol. 422).

Research output: Chapter in Book/Report/Conference proceeding › Chapter › Academic › peer-review

TY - CHAP

T1 - Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data

AU - Koopman, Letty

PY - 2023

Y1 - 2023

N2 - Investigating model fit is essential for determining measurement properties of tests and questionnaires. Mokken scale analysis (MSA) consists of a selection of methods to investigate the fit of nonparametric item response theory models. Existing MSA methods assume a simple random sample, which is violated in two-level test data (i.e., test data of clustered respondents). This chapter discusses the methods manifest monotonicity, conditional association, and manifest invariant item ordering, and investigates the effect of within-group dependency on the point estimate and variability of their statistics. Results showed that fit statistics may be safely used in the presence of within-group dependency, giving appropriate results for sets of items that either did or did not violate assumptions. Implications for practice are discussed.

AB - Investigating model fit is essential for determining measurement properties of tests and questionnaires. Mokken scale analysis (MSA) consists of a selection of methods to investigate the fit of nonparametric item response theory models. Existing MSA methods assume a simple random sample, which is violated in two-level test data (i.e., test data of clustered respondents). This chapter discusses the methods manifest monotonicity, conditional association, and manifest invariant item ordering, and investigates the effect of within-group dependency on the point estimate and variability of their statistics. Results showed that fit statistics may be safely used in the presence of within-group dependency, giving appropriate results for sets of items that either did or did not violate assumptions. Implications for practice are discussed.

U2 - 10.1007/978-3-031-27781-8_20

DO - 10.1007/978-3-031-27781-8_20

M3 - Chapter

SN - 978-3-031-27780-1

SN - 978-3-031-27783-2

T3 - Springer Proceedings in Mathematics & Statistics

SP - 221

EP - 232

BT - Quantitative Psychology

A2 - Wiberg, Marie

A2 - Molenaar, Dylan

A2 - González, Jorge

A2 - Kim, Jee-Seon

A2 - Hwang, Heungsun

PB - Springer

ER -

Koopman L. Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data. In Wiberg M, Molenaar D, González J, Kim JS, Hwang H, editors, Quantitative Psychology: The 87th Annual Meeting of the Psychometric Society, Bologna, Italy, 2022. Springer. 2023. p. 221-232. (Springer Proceedings in Mathematics & Statistics). doi: 10.1007/978-3-031-27781-8_20

Effect of Within-Group Dependency on Fit Statistics in Mokken Scale Analysis in the Presence of Two-Level Test Data

Abstract

Publication series

Access to Document

Handle.net

Fingerprint

Cite this