Generalized linear and nonlinear regression models for DIF detection in longitudinal designs with binary and polytomous responses

Authors: Patrícia Martinková
Publication year: 2023
Publication type: Abstrakt
Periodical: International Meeting on Detecting Differential Item Functioning in Polytomous IRT Models and/or Multiple Groups Programme
Projects: Theoretical foundations of computational psychometrics
Bibliographic citation: MARTINKOVÁ, Patrícia. Generalized linear and nonlinear regression models for DIF detection in longitudinal designs with binary and polytomous responses. In: International Meeting on Detecting Differential Item Functioning in Polytomous IRT Models and/or Multiple Groups Programme. Zurich, 2023, S. 7-7.
ID: 579655

Abstract

ZÁKLADNÍ ÚDAJE: International Meeting on Detecting Differential Item Functioning in Polytomous IRT Models and/or Multiple Groups Programme. Zurich, 2023. s. 7-7. [International Meeting on Detecting Differential Item Functioning in Polytomous IRT Models and/or Multiple Groups 2023. 12.06.2023-13.06.2023, Zurich]. ABSTRAKT: In this work we focus on modelling item responses in the framework of generalized logistic and non- linear models (GLNM). We first discuss a step-by-step development of item response theory (IRT) models via empirical characteristic curves and GLNM, while emphasizing the didactic value of such an approach (Martinkova & Hladka, 2023). We outline possible further uses of GLNM in analyzing criterion-related item validity, detection of differential item functioning (DIF) and differential item functioning in change (DIF-C, Martinkova et al., 2020) suitable for longitudinal designs with binary as well as polytomous responses (Kolek et al., 2022). Finally, we discuss some novel approaches to parameter estimation in the GLNM framework and their practical implementation in the difNLR and ShinyItem Analysis package.