Bayes factor of model selection validates FLMP

Area of Science:

• Cognitive Science
• Computational Psychology
• Perception Research

Background:

• The fuzzy logical model of perception (FLMP) has demonstrated success in describing performance across diverse domains and individuals.
• A key question is whether FLMP's descriptive power is theoretically informative or merely reflects its flexibility.

Purpose of the Study:

• To evaluate the fuzzy logical model of perception (FLMP) using Bayes factor analysis in addition to root mean square deviation (RMSD).
• To investigate the robustness of model recovery under varying experimental conditions and variability.

Main Methods:

• Applied Bayes factor analysis to compare FLMP against alternative models.
• Conducted simulations to assess the ability of RMSD to recover the correct model under different noise levels.
• Examined the theoretical justification and utility of RMSD in model testing.

Main Results:

• FLMP demonstrated a significant descriptive advantage when evaluated using Bayes factor.
• RMSD accurately identified the correct model in simulations, even with added variability.
• The study supports the theoretical justification and meaningfulness of RMSD as a goodness-of-fit index.

Conclusions:

• FLMP maintains its descriptive advantage, supported by both RMSD and Bayes factor criteria.
• RMSD is a theoretically sound and practically useful metric for model evaluation.
• Model selection should employ multiple criteria, analogous to perceptual judgment, rather than relying on a single measure or assuming invariance across parameter values.