Heuristic Assembly of a Classification Multistage Test with Testlets

Zhuoran, Wang; Liu, Ying; Wothke, Werner

doi:10.1007/978-3-030-01310-3_11

Cited by 1 publication

(1 citation statement)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The number of categories and type of items in use also affect the ACT algorithm. Multiclassification that classifies examinees into more than two categories is considered along each dimension, since it is more general than the simpler yet more intensively studied mastery testing (classifying examinees into only two categories) and is prevalent in educational (Wang et al, 2017), occupational (Huebner & Fina, 2015), and medical applications (Cohen, 2008). The number of cutoff scores can vary among dimensions.…”

Section: Introductionmentioning

confidence: 99%

Termination Criteria for Grid Multiclassification Adaptive Testing With Multidimensional Polytomous Items

Wang¹,

Wang

Weiss

2022

Applied Psychological Measurement

Self Cite

View full text Add to dashboard Cite

Adaptive classification testing (ACT) is a variation of computerized adaptive testing (CAT) that is developed to efficiently classify examinees into multiple groups based on predetermined cutoffs. In multidimensional multiclassification (i.e., more than two categories exist along each dimension), grid classification is proposed to classify each examinee into one of the grids encircled by cutoffs (lines/surfaces) along different dimensions so as to provide clearer information regarding an examinee’s relative standing along each dimension and facilitate subsequent treatment and intervention. In this article, the sequential probability ratio test (SPRT) and confidence interval method were implemented in the grid multiclassification ACT. In addition, two new termination criteria, the grid classification generalized likelihood ratio (GGLR) and simplified grid classification generalized likelihood ratio were proposed for grid multiclassification ACT. Simulation studies, using a simulated item bank, and a real item bank with polytomous multidimensional items, show that grid multiclassification ACT is more efficient than classification based on measurement CAT that focuses on trait estimate precision. In the context of a high-quality bank, GGLR was found to most efficiently terminate the grid multiclassification ACT and classify examinees.

show abstract

Section: Introductionmentioning

confidence: 99%