Generalized linear mixed-effects models for studies using different sets of stimuli across conditions

Area of Science:

• Psychometrics
• Statistical Modeling
• Experimental Design

Background:

• Non-repeated item (NRI) designs are crucial in experimental research.
• Existing generalized linear mixed-effects models (GLMMs) may not fully capture dependencies in NRI designs.
• Accurate statistical analysis is vital for valid experimental conclusions.

Purpose of the Study:

• To propose an enhanced GLMM incorporating a level-specific item random effect for NRI designs.
• To evaluate the hypothesis testing performance of the proposed GLMM.
• To compare the proposed model against existing GLMM specifications for NRI data.

Main Methods:

• Development of a novel GLMM with a level-specific item random effect.
• Conducting a simulation study to assess model performance.
• Evaluating hypothesis testing power for detecting experimental condition effects.

Main Results:

• The proposed GLMM with a level-specific item random effect demonstrated superior power compared to existing models.
• Performance improvement was particularly notable when item effect variances were heterogeneous.
• The simulation results support the utility of the enhanced model for NRI designs.

Conclusions:

• The proposed GLMM with a level-specific item random effect is recommended for analyzing data from NRI designs.
• Researchers should consider implementing this enhanced model to improve statistical power and accuracy.
• This advancement offers a more robust analytical approach for experimental data.