Estimation In Generalized Logistic Regression Models For DIF Detection

Autoři: Mgr. Adéla Hladká, Ph.D.
doc. RNDr. Patrícia Martinková, Ph.D.
Ing. Marek Brabec, PhD.
Rok vydání: 2021
Typ publikace: Abstrakt
Projekty: Teoretické základy výpočetní psychometrie
Bibliografická citace: HLADKÁ, Adéla, MARTINKOVÁ, Patrícia, BRABEC, Marek. Estimation In Generalized Logistic Regression Models For DIF Detection [pr]. Virtual, IMPS 2021: The Annual Meeting of the Psychometric Society, 20.07.2021-23.07.2021.
ID: 570044

Abstrakt

ZÁKLADNÍ ÚDAJE: Estimation In Generalized Logistic Regression Models For DIF Detection. [IMPS 2021: The Annual Meeting of the Psychometric Society. Virtual, 20.07.2021-23.07.2021]. ABSTRAKT: Generalized logistic regression models are extensions of logistic regression method for differential item functioning (DIF) detection among binary data which account for possibility of guessing or inattention when responding. In this talk we will discuss several approaches to estimate item parameters including nonlinear least squares, maximum likelihood method and a newly implemented expectation-maximization algorithm. We will further propose a new algorithm based on parametric link function. Differences in estimation procedures will be illustrated with a simulation study and we will also show their implementation in the statistical software R including its package difNLR (Hladká & Martinková, 2020).