Components Identification in Finite Mixture Model

Finite Mixture Model (FMM) are widely utilized in statistical analysis to model unobserved heterogeneity within complex datasets. A key challenge in their implementation is selecting the optimal number of components, as this directly affects the model’s interpretability, stability, and forecasting power. This study focuses on component selection strategies in FMMs, with particular emphasis on two widely used information criteria: the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC). While AIC favors model flexibility, BIC emphasizes parsimony and often performs better in finite samples. We apply the FMM framework to monthly data on China’s rice trade, investigating the relationship among exchange rates, import/export prices, and trade volumes. Empirical findings suggest that a two-component model offers the best fit according to the BIC, revealing distinct market regimes likely shaped by structural conditions and policy interventions.