Bayesian MINBM: addressing multiple inflated values in discrete data

초록

This study addresses the challenge of modeling discrete data with multiple inflated values, where certain outcomes occur more frequently than expected under standard distributions. While previous research has primarily focused on zero-inflated data, the presence of multiple inflated values necessitates more flexible modeling approaches. To this end, we propose the Multiple-Inflated Negative Binomial Mixed-Effect (MINBM) model for analyzing count panel data exhibiting multiple inflated outcomes. Compared to conventional zero-inflated mixed-effect models, the MINBM model more effectively captures individual-level correlations in panel data. By using women’s menstrual cycle data from Kangbuk Samsung Hospital, we demonstrate its usefulness and provide a quantitative evaluation of the mental and physiological factors influencing menstrual cycle length.

키워드