Latent Variable Modeling and Adaptive Testing for Experimental Cognitive Psychopathology Research.

Area of Science:

• Neuroscience
• Psychology
• Biostatistics

Background:

• Cognitive tasks are crucial for assessing neurocognitive outcomes in psychiatric and neurological research.
• Existing cognitive tests often lack strong psychometric properties, hindering their application in clinical settings.
• Computerized adaptive testing (CAT) offers a solution but typically requires large sample sizes and item precalibration.

Purpose of the Study:

• To present a generalized latent variable model enabling adaptive testing without large samples or item precalibration.
• To demonstrate the utility of this model in cognitive testing research.
• To improve the efficiency and precision of neurocognitive outcome measures.

Main Methods:

• Developed a generalized latent variable model integrating mathematical cognitive models.
• Applied the model to N-back task data from a diverse participant sample.
• Conducted a simulation study comparing adaptive versus nonadaptive testing.

Main Results:

• The proposed model allows for adaptive testing with smaller sample sizes.
• Computerized adaptive testing demonstrated a 36% increase in task efficiency.
• Score estimates were 23% more precise with adaptive testing compared to nonadaptive methods.

Conclusions:

• Latent variable modeling and adaptive testing are feasible for experimental cognitive testing, even with limited sample sizes.
• This approach can enhance the impact and replicability of findings in translational studies and clinical trials.
• Adaptive testing holds significant potential for improving neurocognitive outcome measures in neurological and psychiatric research.