Improving test efficiency for a grid multidimensional computerized classification test by the application of a...

Area of Science:

• Psychometrics
• Educational Measurement
• Computerized Adaptive Testing

Background:

• Multidimensional computerized adaptive testing (MCAT) can improve efficiency by accounting for dimensional correlations.
• Previous multidimensional computerized classification tests (MCCT) showed negligible gains over unidimensional tests due to limited item usage.
• Existing MCCT methods using sequential probability ratio tests (SPRT) do not fully leverage inter-dimensional correlations for classification.

Purpose of the Study:

• To enhance the measurement efficiency of MCCT by utilizing correlations between dimensions.
• To improve latent-trait estimation and item selection in MCCT by incorporating inter-dimensional information.
• To evaluate a novel MCCT method using conditional distributions and all administered items for likelihood ratio calculation.

Main Methods:

• Developed a new MCCT approach using conditional distributions of latent-trait estimates.
• Included all administered items in the likelihood ratio calculation within the SPRT framework.
• Evaluated the proposed method through extensive simulation studies.

Main Results:

• The proposed MCCT method significantly improves measurement efficiency compared to standard SPRT.
• Test length reductions ranged from 1% to 32% when dimensions were moderately to highly correlated.
• The method effectively utilizes inter-dimensional correlations, unlike previous approaches.

Conclusions:

• The novel MCCT method offers substantial gains in measurement efficiency by leveraging dimensional correlations.
• This approach provides a more effective way to conduct grid classification in MCCT.
• Findings suggest broader applications for improving adaptive testing accuracy and efficiency.