Tweedie Compound Poisson Models with Covariate-Dependent Random Effects for Multilevel Semicontinuous Data

Area of Science:

• Statistics
• Biostatistics
• Econometrics

Background:

• Multilevel semicontinuous data are prevalent across various fields, including medicine, environment, insurance, and finance.
• Traditional models often use covariate-independent random effects, potentially leading to ecological fallacy and biased results by ignoring covariate dependencies.
• Existing methods fail to adequately incorporate covariates at different hierarchical levels within the data structure.

Purpose of the Study:

• To propose a novel statistical framework, the Tweedie compound Poisson model with covariate-dependent random effects, for analyzing multilevel semicontinuous data.
• To accurately incorporate covariates at their relevant hierarchical levels within the statistical model.
• To provide a method that mitigates the risk of ecological fallacy and enhances the reliability of study findings.

Main Methods:

• Development of a Tweedie compound Poisson model that explicitly accounts for covariate-dependent random effects.
• Incorporation of multilevel covariates at their respective levels within the proposed model.
• Estimation based on the best linear unbiased predictor (BLUP) of random effects, providing explicit expressions for computation and interpretation.

Main Results:

• The proposed model successfully analyzes multilevel semicontinuous data by integrating covariate dependencies.
• Explicit random effect predictors simplify the computation and enhance the interpretability of the model.
• Illustrative analysis of the basic symptoms inventory study data and supporting simulation studies demonstrate the methodology's effectiveness.

Conclusions:

• The covariate-dependent Tweedie compound Poisson model offers a robust approach for analyzing complex multilevel semicontinuous data.
• This methodology addresses limitations of traditional models by accounting for covariate-specific random effects and hierarchical covariate incorporation.
• The developed approach provides more accurate and interpretable results, reducing the potential for misleading conclusions in diverse research areas.