A Markov Mixed-Effect Multinomial Logistic Regression Model for Nominal Repeated Measures with an Application to...

Area of Science:

• Psycholinguistics
• Cognitive Sciences
• Computational Linguistics

Background:

• Syntactic priming effects are crucial for understanding language production and comprehension.
• Distinguishing between activation-based and expectation-based theories of syntactic priming remains a challenge.
• Existing statistical models are insufficient for analyzing repeated measures from multiple participants and items in syntactic priming studies.

Purpose of the Study:

• To develop and present a novel statistical model for investigating syntactic self-priming effects.
• To adjudicate between activation-based and expectation-based theories of syntactic priming.
• To analyze average and participant-specific deviations in syntactic self-priming.

Main Methods:

• A Markov mixed-effect multinomial logistic regression model was developed.
• The model incorporates fixed and random effects for own-category and cross-category lags.
• Category-specific crossed random effects (person and item) were included in a multivariate structure.

Main Results:

• The model provides a statistical framework for analyzing syntactic priming with repeated measures.
• Results indicate that syntactic self-priming effects are consistent with activation-based theories.
• Bayesian analysis was used to evaluate parameter estimate accuracy and precision via simulation.

Conclusions:

• The proposed Markov mixed-effect model effectively analyzes syntactic self-priming.
• Empirical evidence supports activation-based theories over expectation-based theories for syntactic priming.
• The model offers a valuable tool for psycholinguistic research on language production mechanisms.