Return to search

Lasso Regularization for DIF Detection in Graded Response Models

Previous research has tested the lasso method for DIF detection in dichotomous items, but limited research is available on this technique for polytomous items. This simulation study compares the lasso method to hybrid ordinal logistic regression to test performance in terms of TP and FP rates when considering sample size, test length, number of response categories, group balance, DIF proportion, and DIF magnitude. Results showed better Type I error control with the lasso, with smaller sample sizes, unbalanced groups, and weak DIF. The lasso also exhibited more stable Type I error control when DIF was weak, and groups were unbalanced. Lastly, low DIF proportion contributed to better Type I error control and higher TP rates with both methods.

Identiferoai:union.ndltd.org:unt.edu/info:ark/67531/metadc2137561
Date05 1900
CreatorsAvila Alejo, Denisse
ContributorsHull, Darrell, Uanhoro, James, Boesch, Miriam C., Acar, Selcuk
PublisherUniversity of North Texas
Source SetsUniversity of North Texas
LanguageEnglish
Detected LanguageEnglish
TypeThesis or Dissertation
FormatText
RightsPublic, Avila Alejo, Denisse, Copyright, Copyright is held by the author, unless otherwise noted. All rights Reserved.

Page generated in 0.002 seconds