A note on the applied use of MDL approximations

Area of Science:

• Cognitive Psychology
• Psychometrics
• Information Theory

Background:

• Understanding and modeling psychological retention is crucial for learning and memory research.
• Nested models are frequently used in psychology, but their comparative complexity can be challenging to assess.
• Minimum Description Length (MDL) offers a principled approach to model selection based on data compression.

Purpose of the Study:

• To compare two nested psychological models of retention.
• To evaluate the performance of the Minimum Description Length (MDL) principle in model selection for psychological data.
• To investigate the complexity of nested models using information-theoretic measures.

Main Methods:

• Application of the Minimum Description Length (MDL) principle for model comparison.
• Calculation of the normalized maximum likelihood using the Fisher information approximation.
• Comparative analysis of two nested psychological retention models.

Main Results:

• The full psychological retention model was assigned a smaller complexity value by the MDL criterion.
• This result holds even for moderately large sample sizes.
• A geometric interpretation of the complexity assignment was explored.

Conclusions:

• The MDL principle favors more complex models in this specific psychological retention context, contrary to some expectations.
• The findings have practical implications for selecting appropriate psychological models in research and application.
• Further geometric and theoretical exploration can illuminate model selection dynamics in psychology.